You can subscribe to our announcement list at:
http://groups.yahoo.com/group/AdaptiveAI/join

• 2006
• 2005
• 2004
• 2003
• 2002
• 2001

July 19, 2007

More Good News

I recently returned from visits to Australia and England where I brought some of our key investors up to date. These trips resulted in additional investments in our company that will allow us to gear up to about 24 full-time staff over the next 12 months or so. We are currently actively seeking at least one more high-level programmer (see Opportunities).

Our team just gained two bright, highly motivated new members (see Team)

Matt Johnson is an AGI theorist, as well as an experienced programmer. Matt is working with me on various cognitive design improvements, including our system's knowledge representation. We expect him to make significant design contributions to the ongoing enhancement of our AGI.

In the few weeks that Chandra Barnett has been with us she has already tackled a comprehensive upgrade to our core engine that improves its modularity and scalability. Chandra is also very active in understanding, and contributing to, cognitive aspects of our project.

Given the staff and additional funding we are confident that we can develop our AGI technology to a point of significant practical utility over the next 18 months.

Again: Exciting times!

Peter

April 2, 2007

A Growing, and Stronger Team.

Over the past few months our team has grown from 8 to 13 full-time staff.

Two months ago, Mark Ekstrand joined us to manage our data, training, and testing team. Mark has extensive experience in quality assurance and project mangement, and brings significant practical wisdom to our project. His efforts have already put us in a much better position to handle the hundreds of implementation tasks that face our data department.

Kurt Rodgers joined our data department a few weeks ago to help compile and organize our training data. Kurt's passion for language theory is also finding a clear outlet in his involment with our NLP (natural language processing) and ontology development.

The rest of our team is also doing very well as the excitement of getting closer to our project goal is mounting.

I am also very pleased to annouce the hiring of an additional three experienced programmers with particular passion for AGI. These three individuals will start working with us within the next 2 months, and we'll announce details at that time. This brings our total of full-time employees to 13.

Our new premises are working out very well, while providing room for further expansion.

With an additional investment from our largest investor, our finances are still in good shape. We have sufficient cash on hand for our existing team till the completion of 'Project Aigo' - the end of next year.

At this time our focus is on a May 2007 milestone, when we expect to showcase our progress to various investors.

Exciting times!

Peter Voss

December 21st, 2006

Project Aigo' is progressing well. We have good news to report on a number of fronts

Staff - Four more full-time members have joined our team since the last update.

A few months ago, Dean Gores moved from Florida to join a2i2. Since then he has proved himself to be a highly productive programmer, adding significant functionality to various aspects of our system.

Andy Lech recently joined us as an experienced programmer. Andy has already made important contributions to various algorithms and tools.

Next, we welcomed back our very first employee, Steve Dotson, in the capacity of AI psychologist. After pursuing other interests for a few years, Steve is returned from the East Coast to help us train and test our system, as well as helping out on the administrative side.

Finally, I'm happy to have finally filled the key position of director of engineering: Basalat Raja, a highly experienced software engineer has joined us to head up software engineering. Basalat's skills, motivation, and personal philosophy make him a welcome partner to our venture.

Please see bio page for more on these new team members.

At this stage we are only actively looking for two more staff members, including: Chief AI Psychologist. This crucial senior position entails managing our knowledge acquisition, training and testing efforts. See Opportunities.

Premises - In order to accommodate our current staff expansion as well as projected growth, we have acquired additional premises nearby. In total we now have enough office space for about 20 people. We expect this to suffice for the next 18 months or so - almost to the completion of 'Project Aigo'.

Technical progress - While we do not publicize many details of our business plan and technology, I can report that our project is on schedule. The staff shortage we experienced over the past year has been largely offset by improved efficiencies and various development shortcuts we discovered. As a key milestone, we just implemented a version of our prototype to do useful work for our project, assisting us in various simple tasks.

Finances - Various cost savings, combined with some additional investment received, put us in a good financial position. We have sufficient funds for the next major phase of the project.

August 03, 2006

News Flash

Our project is progressing well, and we are once again looking to expand our team.

This is an opportunity for a select few individuals to become members of our core team, and to significantly contribute to the creation of real AGI.

Various Los Angeles based full-time positions are currently open. All offer competitive cash compensation.

Please see http://adaptiveai.com/company/opportunities.htm  for details, and apply to jobs@adaptiveai.com

Peter Voss

May 09, 2006

Our Team Grows

Our recruiting drive has finally paid off -- I'm happy to announce that we have three additional full-time members on our team.

Don Kim, an experienced programmer, has joined in the capacity of AI Psychologist and tools developer.

John Burge, a highly experienced AI, Cognitive Engineering, and Natural Language Processing researcher, has taken charge of our systems training, testing, and data acquisition needs.

Philip 'Flippie' van Eeden, a brilliant software guru, is taking over many of the trickiest coding and code design tasks from me. This is freeing me up to spend more time on fundamental cognitive and architectural development, and well as on overall project management. Flippie previously implemented the low-level aspects (including interpreter and compiler) of a comprehensive programming language I developed in the early '90s.

Please see bio page for more on these new team members.

With these additions, we are not actively looking for additional staff right now, but do have another one or two slots open for people with a very good fit. We will identify our new staffing bottlenecks over the coming months.

Project on Track

Even with the many months of being short-staffed we have managed to make good technical progress. Given the strength of our current team, plus our healthy financial situation, I expect that we will be able to complete 'Project Aigo' on time - late 2008.

Our focus is now firmly on making our AGI smarter - and fortunately now have few distractions.

Peter Voss

March 08, 2006

November 8th, 2005

Firing on all cylinders - and needing more...

Since our last newsletter two new team members have joined us permanently: John de la Garza as programmer, and David Raw as AI psychologist (their profiles are on the TEAM page). They've both come up to speed and have made solid contributions. In fact, David's expertise has already progressed to a point where we are ready to replace his entry-level position.

All in all, our team is firing on all cylinders, and we've made very gratifying progress in implementing core functionality of our commercial prototype. Having secured adequate funding for the current phase of our project, our bottleneck is now firmly in finding 'more cylinders'.

Over the past few months we've actively interviewed to fill the team leader (CTO candidate) position, but have only managed a few 'near misses'. We are still looking for the right mix of technical and management ability, plus the desire and ability to get immersed in our project.

At this stage we are looking for 4 additional people who want to be involved with actually creating 'real AI'. With immediate effect we have openings for the following LA-based, full-time positions:

• Senior Programmer/ Team Leader 
• Chief AI Psychologist
• Experienced Programmer
  
• Entry-level AI Psychologist/Programmer.

All positions offer a highly stimulating work environment, market-competitive compensation, generous equity participation, and a chance to become core members of our rapidly growing company. A preference is given to those who truly want to share in our vision, risk, and reward.

July 3, 2005

All Systems Go for Project Aigo

Details of our business and technical development plans have become a lot clearer over the past few months. This knowledge, together with several other developments and insights, have prompted us to initiate an accelerated implementation plan.

This updated plan for Project Aigo sees us adding about three additional team members immediately, and concentrating almost all of our efforts for the next 10 months or so on implementing key functionality of our new system. This approach, in contrast to our original plan of substantially gearing up in the near term, has a number of advantages. These include having the much greater efficiency of a small team, and fewer delays and distractions caused by recruitment, fund-raising and training. It also provides additional time both to find a suitable CTO and to build relationships with additional financial partners.

This current, highly productive phase is the next major step towards completion of 'Project Aigo': fully functional, commercializable AGI technology.

Thanks mainly to a large additional investment by one of our existing shareholders, Norman Pater, we are now fully funded for this phase.

More good news is that recent planning and development work have prompted a downwards revision of estimates for both cost and time required to complete this project. We are now aiming for August 2008. As per our business plan, at that point we expect our company to transition from (almost) pure technological development focus to product development, production, and marketing.

In the meantime though, we have a lot of work to do!

Additional Team Members

With immediate effect we have openings for the following LA-based, full-time positions:

1) Experienced programmer/ team leader (potential CTO)

2) Chief AI Psychologist

All these positions offer a highly stimulating work environment, generous equity participation, and a chance to become a core members of a team creating 'real AI'. While cash compensation is negotiable up to market rates, a preference is given to those who truly want to share in our vision, risk, and reward.

Partnerships and Investments

While we are not actively seeking any funds at this time, we are very interested in building relationships with potential strategic business partners and investors.

April 6, 2005

Funding Momentum

We've had a great start to our funding drive: One of our existing investors has decided to increase his investment by half a million dollars!

This vote of confidence will help us raise the additional several million dollars we need now to gear up our project sufficiently to meet our ambitious targets. In the short term, this cash injection allows us to start recruiting key personnel - beginning with the most important position: a CTO (see below).

Our project plan and demo has now been presented to the first few potential investors, and we have received favorable feedback. As we approach additional partners we continue to flesh out and hone project details.

We are still seeking potential advisors and/ or business partners - visionaries who want to be part of accelerating this breakthrough and to help shape its destiny. In presenting our vision and potential, I am constantly reminded of how profoundly AGI technology promises to change our world - enabling life-enhancing progress in just about all areas humans care about. Please contact me if you know anyone who may want to get involved with our project.

Technical Progress

Our new AGI framework's core structure and initial set of probes are in place and functioning. Our clean-slate redesign is showing good promise: some crucial basic operations are several orders of magnitude faster than in our previous proof-of-concept prototype.

We have come up with a tentative roadmap for implementing the cognitive capabilities our system requires, and started work on the first few items. Over the next few months we expect to re-implement most of the features from our previous research phase in this new 'industrial strength' AGI system. In addition, we are busy implementing a number of new, higher-level learning/ thinking abilities.

Unfortunately, for proprietary reasons, we are not able to offer much technical detail about our work at this stage.

Search for CTO, and other Opportunities

I am now actively seeking a dynamic manager to build and manage a team of engineers and scientists. This hands-on, technically competent individual will be someone wanting to become immersed in the realization of our vision - and desire to be a key partner in our venture and success. The ability to rapidly learn and understand a wide range of technologies, together with knack for selecting and motivating bright individuals are core requirements. A passion for creating 'real AI' is crucial, specific AI experience is not. A highly negotiable package will include a significant equity component. For detail see CTO position.

If you know anyone who may fit this profile, please pass on this information or let me know.

While it may take a while to find a CTO and bring him up to speed, we also have an immediate opening for a full-time, LA-based programmer. See Programmer position.

As a reminder: We now have a company email list for new opportunities and contract work requests: http://groups.yahoo.com/group/a2i2Work/join . Please join if you may be interested in working on our project.

The people who are already working on our project,  now have a page, detailing current members.

Peter Voss

Towards Increased Intelligence!

March 1, 2005

Project Aigo 

We have been busy.

Since deciding just over two months ago to gear up our project to actually building a fully functional, high-level AGI within 4 years (Project Aigo), we have made some good progress:

  ·             We have assembled a group of business and futurist advisors, some of whom are profiled here. Brainstorming our project plan with them has already provided several key insights, as well as some new important contacts.

  ·             Our financial flexibility has been further improved by two new (modest) investments. These additional funds will allow us to immediately take on another one or two people.

  ·             We have just completed an initial version of our demo. This demo will help us communicate our vision of the power and flexibility of our planned AGI system. 

  ·             We have started approaching a number of potential financial advisors/ partners, and plan to present our demo and project plan to them over the coming weeks. 

Looking ahead, we have lot more work to do on fleshing out our project plans, both technical and commercial. We also expect to further enhance our demo.

Please contact me with suggestions for additional advisors and/ or business partners (for details see previous newsletter).

Looking Ahead

A result of our recent project planning has been the realization that we really don't want the grass to grow under our feet while we look for financial partners. With this in mind we have left behind our 3-year-old research prototype and started work on the new framework required for Project Aigo. With a bit of luck (and a lot of work and skill) we expect to integrate our new AGI with our demo within a few months.

Opportunities

In another initiative, I have just created a new company email list for new opportunities and contract work requests. Please join if you may be interested in working on our project now or in the future.

We will be posting details of three potential key management positions in our company: general management, technical, and (AI) psychology.

In the short term, we are looking for additional AI psychologists plus a part-time marketing assistant.

Now back to work...

Peter Voss

January 7, 2005

The Next Phase

I'm happy to report that the last three years' efforts have propelled our project to a point where a new, bold initiative is warranted.

Having satisfied ourselves that we've cracked the key issues required for building a system with general intelligence, we are ready to move beyond our current exploratory (internal) proof-of-concept prototype, and to embark on the extremely ambitious task of actually building a fully functional, high-level AGI. We estimate that this will require a team of up to 50 people, and about four years.

In order to properly showcase our approach and technology to future strategic partners, we are dedicating the next several months to the development and construction of a demonstration system. We are currently working on its specification.

Seeking Advisors

In order to maximize our company's potential, and to minimize risks, we are now actively seeking a small group of dedicated (but part-time) strategic advisors. I am looking to assemble a permanent board of smart, yet practical futurists people with an appreciation of the implications of full-blown AGI. Ideally, though not essential, candidates would also have high-level business experience. AI background is optional.

This panel will help to shape the difficult decisions ahead: Attracting and structuring appropriate equity partnerships, managing company growth, deciding on applications and commercialization, as well as brainstorming IP and technology implications.

Please let me know if you know of anyone who might fit the profile (and thank you to those of you already advising me).

Other News

Our team has remained stable over the past few months, we have successfully completed our move to our new larger 'AI Kommune', and are continuing to be quite productive. We are all very excited about our new plans and look forward to tackling our new challenges.

Current financial reserves will very comfortably see us through this interim phase; in fact, we hope to have one or two additions to our team in the near future. We are also considering outsourcing some specific aspects of our work. We'll publish any opportunities as and when they arise.

Peter Voss

October 21, 2004

Brainpower

I'm happy to report that we now have what looks like a stable six-fulltime-person, LA-based team. Our latest addition is Reza Assadi, who joined us as AI Psychologist/ Software Engineer from Canada a few weeks ago (yes- he actually managed to get a visa!). Reza has a strong background in electronics, mathematics, computer science and programming, and has already contributed some good ideas and work to our project.

Our team is the strongest it has been, and a pleasure to work (and live) with. At this time we are still looking for one more LA-based AI Psychologist (see Opportunities).

Productivity has been pretty high in the past few months, but is unfortunately currently being undermined by an unscheduled move to new digs. On the up side, we are about to move to a very pleasant, nearby house that provides quite a bit more space.

Technical Progress

Our work towards Phase 3 (see previous news) is now at a point where we believe we have 'thin' (in some cases very thin!) implementations of all major cognitive features required. Our focus now is threefold: 1) More fully integrating the various abilities. 2) Improving ability and performance. 3) Developing more sophisticated and challenging test environments.

One key development is a newly devised innovative training system. This, in conjunction with a rudimentary 'language' layer will allow us much more rapidly to test complex sets of knowledge/skills acquisition. We expect this work to culminate in a highly integrated proof-of-concept prototype of high-level animal intelligence over the next 6 to 8 months - an ambitious goal.

The Future

Recent brainstorming by the team is starting to address issues beyond Phase 3 - we are planning integration of infant language acquisition and ongoing, cumulative learning. Specifically, we are looking at requirements of being 'ready-to-learn' (in the pedagogical sense) in order to bootstrap self-motivated knowledge/skill acquisition.

On the organizational front, we are still in fairly good financial shape with existing funding till at least the end of next year (at our current shoestring level). However, more ambitiously, subject to successful completion of our Phase 3 proof-of-concept demo, we now plan to team up with a suitable strategic (angel) investor by roughly the middle of 2005. This will enable us to speed-up progress substantially on achieving our ultimate goal: High-level AGI.

Finally, for those of you attending the AC2004 Conference in Palo Alto in early November (5th-7th), Josh, Justin, Todd, Louise, and I will be there - so we hope to see you.

July 6, 2004

Moving Forward

The A2I2 team

March 13, 2004

(Human) Brain Power

Since the last update we've been fortunate to find and integrate three new members into our team. Two very smart, young AI Psychologists moved into our "AI Kommune" at the end of January. Justin Corwin hails from Utah, while Josh Truett came all the way from Virginia. Both had been doing independent research into AGI for some time while monitoring a2i2's progress. Responding to our recent advertisement they decided to contact us. During short getting-to-know-each-other visits to LA it didn't take us long to conclude that joining forces would be mutually beneficial - there was a great fit at technical, motivational and philosophical levels.

The third new member of our team is Todd Rockhold, a highly experienced programmer, who has started handling our framework and tools programming. Todd was also especially attracted to our project by its ambitious futurist goals and our specific approach to AI. All of our team members are now based in LA or within reasonable driving distance.

On the downside, we are about to lose Richard from our team - due to unfortunate personal circumstances he has to return to Seattle :-(

We hope that he'll be able to rejoin us sometime in the future. Richard has made valuable contributions in streamlining our automated testing capabilities, in cognitive research, and in managing our network and documentation. All the best, Richard!

Phase-3

The addition of our three new members prompted us to re-assess the degree of Phase-3 planning we should do. We decided that there were substantial benefits to developing a thorough specification of our next major milestone. We ended up investing about 6 weeks in researching, brainstorming, and writing up the desired functionality for Phase-3. This turned out to be a good investment as a training exercise, and more importantly, to give a much clearer vision of where we are heading.

Phase-3 is ambitious: In addition to substantially upgrading our systems current cognitive abilities, we plan to implement many new features, including basic meta-cognitive control of problem-solving strategies, as well as proto-language and proto-numeracy. Overall, we expect this phase to cover practically all non-instinctual animal mental capabilities at a conceptual, or proof-of-concept, level.

We expect this phase to yield an impressive enough prototype to allow us to raise the more substantial amounts of capital that we will probably need to fund development of a high-performance prototype.

The excellent momentum and direction of our current team, taken together with the fact that we have already secured sufficient funding for this phase, makes us quite optimistic that we are heading into a period of exceptional productivity. Our current estimate is that we'll be able to complete most of the items on our Phase-3 agenda within about 12 months.

Looking for another 'AI Psychologist'

We are currently looking to expand our team by one more person to help with the design, implementation, and evaluation of cognitive tests for our AGI. We are seeking someone with a passionate interest in AGI and theories of cognition. This position requires the integration of insights from various areas of cognitive psychology and computer science. Past working experience or formal qualifications in these fields is not necessary, though programming experience is a definite plus.

We expect candidates to have studied our website, and to be in broad agreement with our general approach.

The position is LA based, and we require attendance at very frequent brainstorming sessions. Compensation is primarily in the form of stock with only subsistence level pay and/or free accommodation. The position is fulltime, but with flexible hours. Work environment is friendly, informal, and intellectually rich. We are generally unable to meet visa requirements for non-residents.

For serious inquiries please email both Peter Voss and Michael Michalchik

Till next time.

Towards Increased Intelligence!

Peter Voss

January 4, 2004

AGI Project Review for 2003

Summary

Last year's efforts rewarded us with excellent overall progress towards the company's current main objective: building and evaluating an AGI (Artificial General Intelligence) proof-of-concept prototype. More specifically, we experienced a major advance in the project - a definite transition from focusing on infrastructure development (AGI framework, tools, core data structures and probes) to expending most of our effort improving the intelligence of our system. The past three months have been particularly productive, showing significant advances in our AGI's cognitive ability, and our testing and evaluation efficiency (see details below). All of these gains were made in spite of spending many frustrating months being acutely short-staffed, requiring the investment of substantial amounts of time recruiting and training.

On the financial front the project is in fairly good shape. I am currently finalizing a round of funding that has already yielded sufficient money to comfortably cover this year's operations (at our current, very modest, expense level). I expect that the next round will be late this year, and that it will be based on the (then) demonstrable strength of our prototype, rather than just on our enthusiasm and reputation, as has been the case up to now.

We are once again actively looking for 2 or 3 additional full-time, LA-based team members to help with the design, coding, testing, and evaluation of cognitive features. Programming experience is a plus, but not essential - a lot of our current work involves dreaming up new tests, configuring test environments, evaluating results, and brainstorming improvements. Core requirements are intelligence, curiosity, and a passion for being involved with our kind of radical AGI development work. A deep interest in cognitive (experimental) psychology and/or cognitive science (theory-of-mind) is highly desirable. Contact:
Peter Voss

Over the past year we've achieved many of the technical objectives we had set ourselves. Indeed, we developed some important features not originally planned for during this period. Particularly pleasing is the range of different tasks that our standard, untrained AGI can already learn autonomously; just by interacting with the (virtual test) environment - without additional programming or human interaction.

In recent months we have made significant breakthroughs in the cognitive sophistication of our system. Starting with a robust system capable of reliable pattern recognition and adaptive stimulus response association we have developed and integrated more advanced cognitive abilities. These include:

• A basic ability to categorize, abstract, prioritize the information entering the brain based on that information's problem solving value. The system can do this without preexisting definitions of the type of information that is important.
• The ability to encode causal relationships in arbitrary environments, and to use them to plan multi-step solutions to reach goals.
• Multifaceted and general exploratory routines. These are able to handle a wide variety of heterogeneous environments that include obstacles, distractions, misdirecting signals, and reward and punishment sources.
• An ability to identify deficits in experience or incomplete exploration, and to take multi-step actions to fill holes in its causal relationship knowledge base.
• A number of temporal processing advances, including features that allow the system to cope with delayed feedback, and to discriminate and act on temporal patterns.
• An initial version of sense memory that allows the system to act on recent experience whether or not it corresponds to overall experience.
• A rudimentary ability to coordinate action and sensory representations across sensory modalities i.e. hand eye coordination.
• Meta-cognitive states and controls (aka emotions) including such things as boredom, disappointment and curiosity.

As pleased as we are with these advances, much more needs to be done. Among our near-term projects are improvements in:

• Conceptualization/ generalization and abstraction.
• Goal creation and activation system.
• Temporal pattern processing and the handling of delayed and conditional causal relationships.
• Encoding and using associations and context.

I'm currently in the process of detailing our next major development objective, Phase-3.

We look forward to a highly productive 2004 during which we hope to significantly elevate our prototype's intelligence, and to solve many of the fundamental problems of achieving Real AI.

Recent

I'm very happy to report that the 'great buzz' reported in the last news bulletin transformed itself into substantial, productive momentum. The whole team chipped in to make this one of the best quarters ever.

In addition to numerous enhancements to our virtual world system, David added an active hand and a more versatile eye option to our AGI's probes. These are being put to good use in our Piagetian inspired 'artificial infant' development path. He also developed the 'SnapShooter' - a configurable 'flight-recorder' tool that helps us retrace brain/ world interactions, and also provides the brain's perspective of the data it perceives.

Temo made important improvements to our virtual world scripting system. We can now easily set up complex changes and interactions of entities/ properties in our test environments. He also developed new schema viewing and topological tracing tools, and was instrumental in coding significantly enhanced exploration routines.

An overhaul of our action selection logic was undertaken by Béla, who also designed and managed our new exploration logic. In addition to that, he took over 'schema' (cause-effect encoding) development from me, and elevated it to its current level of functionality. More recently, he implemented an initial version of short-term-memory of context, a feature central to goal activation.

Richard, in addition to assisting with system admin and logistic tasks, has substantially streamlined our testing efficiency, and now fully manages our automated batch test system. This adds efficient 'black box' testing to our 'white box' (detailed analysis) and 'gray box' (theoretical or virtual analysis) [thanks, Troy] efforts. Richard also helps with the design of new test environments, and evaluates new tools and features.

The brunt of dreaming up and implementing meaningful cognitive tests for our emerging intelligence continues to fall on Michael - who does an excellent job if it. In the past few months he has developed and refined test environments for AIGO (our artificial, virtual dog), 'baby', as well as specific cognitive features. Additionally, through our brainstorming sessions and white-papers, he has been an active and valuable contributor to the design of the AGI itself.

With Michael now routinely spending a few days a week here at the 'AI Kommune', the LA-based team now has regular and extensive brain-storming sessions that are invaluable (and enjoyable!).

My own recent focus has been on overall technical planning, design, and management, as well as the design and implementation of specific (new or improved) cognitive algorithms. In addition to my original 'schema' implementation, I've recently added new temporal pattern recognition routines, perceptual short-term-memory, vastly improved 'conceptualization' logic, proportional output activation, and various improvements to recognition and explore logic.

Unfortunately, both David and Temo have recently gone on (hopefully temporary) hiatus from the project - Mr. Green calling. As we are now looking for additional team-members, we've become extremely aware of the limitations of remote collaboration: Our project moves so fast, and so many issues need to be brainstormed at high bandwidth that it is really imperative that our core team be in one location.

Looking Ahead

As mentioned above, our funding is secure for this year (in a shoestring budget sort of way). In order to capitalize on our financial situation and great current momentum we are looking to increase the size of our team here in LA to about 6 or 7 people - and are in a position to cover (very) basic living expenses in cash and/or by providing accommodation.

Now that most of our time is spent actually increasing our AGI's intelligence, as opposed to building infrastructure and tools, we expect to make rapid and substantial progress in our quest for high-level general artificial intelligence.

Our focus during this year will remain almost exclusively on proof-of-concept functionality rather than near-term practical or commercial applications. Having said this, though, I do expect our current work to yield a very impressive demonstration system of our AGI technology later this year.

Depending on funding, progress and investor prospects, I expect that the strength of our prototype will allow us to raise funds on a favorable basis sometime around the end of the year. I hope that the next round will elevate our project to being well funded, and to implement the next phase; (most likely) a practical, functional prototype with vastly improved performance and capabilities.

The task we've set ourselves is enormously challenging, yet I feel confident that during this year we can make a substantial dent in achieving our long-term goal: creating a practical, high-level general intelligence engine.

Lastly, I'd like to thank everyone who supported our project during 2003, including short and long-term team members, investors (you know who you are) and my partner, Louise Gold, for her support.

Till next time.

Towards Increased Intelligence,

Peter Voss

October 10, 2003

There's a great buzz at the a2i2 Kommune.

After many months of frustrating, time-consuming searches, and several false starts, we are now thrilled to have two new permanent members living at the Kommune and working on the project full-time: Béla and Richard Kowalski.

Béla , apart from being a really interesting and friendly human being (for now, until posthuman technology arrives), is a highly experienced system architect/ programmer with a passion for knowledge - especially AGI. In less than two weeks he has learned enough about our system and C# to be designing and re-coding one of our system's central cognitive algorithms.

Richard , a friend from the transhumanist community, comes to us with practical computer and systems testing experience. He is responsible for setting up, evaluating, reporting on, and designing AGI performance tests - an increasingly crucial aspect of our project from now on. He will also become involved with programming in the near future.

Both are active futurists, extropians, and life-extensionists.

Welcome to both of you - it is a pleasure having you around.

For the past two weeks we've also had Michael joining us here at the Kommune for a few days at a time. His presence has contributed to some particularly productive and enjoyable team-work in recent times.

Next week we'll have David and Temo joining us from Canada for our annual brainstorming pow-wow. We all look forward to this opportunity to catch up, learn, innovate, plan - and have fun.

Daniel Batty , who lived here and helped with our AGI testing for almost three months unfortunately had to return to England (visa problems - what else!). All the best, Dan.

Troy Gardner , our longtime house-mate and a2i2 partner, has been working full-time (in fact, mega overtime!) on another commercial project. One consolation is that he still occasionally treats us to his unique cauliflower sushi (patent pending).

Project Progress

The upshot of our recent staff shortage, together with time taken for recruitment and training, is that progress has been slower than hoped for. However, in spite of those limitations, the project has advanced significantly - and we are quite pleased with many aspects of our AGI's current performance.

Since the last new bulletin we have made tremendous progress in automating our performance testing, and simplifying virtual world setup through the use of a new scripting system. Our virtual world system has also been improved and expanded substantially - both in terms of the number and function of simulated senses and behaviors, and in terms of raw performance and convenience. The new scripting system also allows the system to dynamically and interactively re-configure itself (both the virtual world and the AGI) 'on the fly' - a major advance. Thanks David and Temo !

On the cognitive side, dozens of incremental improvements have enabled our system to now perform well on an increasingly wide range of (simple) tasks. Several significant new 'brain activity' logging and analysis tools are substantially improving our productivity on ever more complex configurations.

Much of our current work is focused on combining many cognitive tasks into one or two highly integrated environments (tests). These setups will test and develop the AGI's ability to cope with conflicting requirements. They also form the backbone of our near-term dual-development-path strategy: one path following the 'artificial animal' model (see AIGO description in previous email), the other, a Piagetian infant development model. The two paths have a lot of overlap - and over time, we expect them to converge.

The current ' Piagetian path ' aims at replicating (very sparsely and abstractly) human infant ability at roughly the 8 months level. To this end we have just completed the first implementation of a major new cognitive ability: 'schemas' - the ability to learn and use probabilistic cause-effect sequences. We are also in the throes of vastly improving the system's basic exploration routines.

We optimistically hope that with our recently strengthened local team we will achieve most of the specified functionality (for both AIGO and ' artificial infant ') by early next year.

At our present burn rate, we currently have sufficient funds for about 8 to 10 months' operation - thanks to the support of various investors. We will be seeking additional capital within the next few months.

As always - if you, or anyone you know, would like to invest in or join our project, please contact me at peter@optimal.org

Till next time.

Towards Increased Intelligence,

Peter Voss

April 28, 2003

Some New Blood, Upgraded AGI Framework, and a New Milestone.

Friends of a2i2,

Lots of activity again since our last communiqué:

Team

Rob 's impending fatherhood has wisely motivated him to seek employment that is less risky than proof-of-concept AI research (on a shoestring budget). Congratulations - and we wish him well. Rob's legacy of thorough component testing will continue to enhance our system's robustness.

On the up side for us, Ascanio Pignatelli , a bright highly-motivated renaissance man with a keen interest in theory of mind, has joined our team full-time. He will be helping Michael and myself train and evaluate our budding AGI system.

David Braginsky , our local software star, has succumbed to an irresistible offer from Google to dramatically improve his monthly cash-flow. We are happy that David stayed with us several months beyond his original commitment, and that he (almost single-handedly) coded the rich repertoire of cognitive algorithms that now form the basis of our system's intelligence. All the best, David!

David McFadzean (AKA '#8') and I have taken over the ongoing task of enhancing our cognitive core.

Joining #8 in Montreal is Temo Chalasani , a programmer with substantial academic and commercial AI experience. Temo will be working closely with David to enhance our development framework, including the Virtual World and various Probes.

Welcome, Ascanio and Temo !

Development

Most of our effort over the past two months has focused on facilitating the evaluation and training of our AGI's cognitive ability. To this end we undertook some major streamlining of our test framework, eliminating obsolete features and unnecessary complexity. This effort has provided a many-fold performance improvement, and has substantially simplified our system. Practically it has sped up programming and debugging by about 25% - in effect giving us another half a programmer :)

#8 recently completed a major upgrade of our Virtual World, adding a multitude of senses and reactions to the world's virtual entities. Michael has been very busy creating test set-ups that exercise the tremendous flexibility of this design. We expect that we will be able to develop significantly intelligent behavior with this tool. Troy, on the other hand, has created a more robust mechanism to transport test data between software versions, and improved the all important visualization tools that allow us to look 'inside the brain'. David B's swansong was the re-development of our context/ temporal priming system - and we have started incorporating this into our tests.

Overall, we are making good, steady progress with our project. Being close to attaining our original phase-1 goals, Michael and I are now working on the detailed specifications of an aggressive phase-2 development schedule. These plans include simulating many 'dog-level' abilities - call him 'AIGO'. Aided by the improvements made to our test framework, our more optimistic estimates are to achieve this milestone within 4 to 5 months.

With the ongoing support of our investors and team members, we anticipate continuing to enjoy a 10 to 12 months window of secured funding - that's good.

Currently we have adequate programming resources, but we could still use additional help here in LA to relieve our testing/ training/ evaluation bottleneck. Any bright, motivated real or wanna -be cognitive scientists out there? Accommodation can be provided. Contact me for further details - peter@optimal.org

Till next time.

Towards Increased Intelligence,

Peter Voss

March 3, 2003

Friends of a2i2,

This time:

1. Meet Michael, 
2. Overview of our technical status, 
3. Another call for additional team members.
4. Plus, we now have another investor supporting our project: Johann Gevers - a futurist and friend - has made an initial contribution, and is increasing his holding on a monthly basis.

1.) I'm happy to announce that Michael Michalchick has joined our team as senior cognitive scientist. Michael spent 2 years in Antonio Damasio's neuroscience lab as part of    his 4 years in the University of Iowa neuroscience Ph.D. program. His experience in cognitive psychology (including psychometric testing, test design, plus raising an infant!) is particularly relevant to our project, where his focus is on developing integrated testing/ training protocols for our emerging AGI. He will also be involved in the ongoing process of designing and improving its mental ability. Michael has already contributed many suggestions in our regular brainstorming sessions and is currently testing various combinations of sense modalities and cognitive algorithms.

2.) Technical Status Review: Over the past two months we've been rather busy - and have made some good progress: We now have a set of seven 'probes' (input/output sense/actuators) that cover both real and virtual world interaction. They include (very elementary) visual and auditory sense inputs, as well as various computer/ user/ virtual world outputs. Our virtual world system has been enhanced with a number of useful simulation and control features, and is currently undergoing another major upgrade (thanks #8 - AKA David M).

Cognitive functionality now integrates sensing, learning/ forgetting, action/ exploration with rudimentary goals, meta-cognition, and 'emotions' sub-systems. Learning modes include: exposure and 'imprinting', reinforcement (incl. 'instrumental conditioning'), and association (incl. a variation of 'classical conditioning'). Pattern recognition utilizes 'fuzzy matching', and supports pattern completion and recall. (Dave B and Rob bore the brunt of developing, testing and enhancing these features.)

Our AGI framework allows substantial flexibility in probe configuration and parameterization. This aids in the exploration and analysis of different data and systems dynamics. We also now have three different structures for automated testing/learning, and a new setup and statistics system, that have significantly improved testing efficiency. A greatly improved network visualization tool (courtesy Troy) helps us peek 'inside the brain'.

Our short-term focus is on further integrating and improving current functionality, and to push development of temporal dynamics (incl. temporal patterns, context, and actions) and conceptualization mechanisms. We also plan to put additional effort into long-term, self-sustaining incremental learning. We have a lot to do. Exciting times!

3.) Additional Team Members: We are once again actively looking for additional members for our team. On the coding side, we are need a moderately to highly experienced programmer to be involved with all aspects of program enhancements and new designs. (We use C# under VS.NET.) Specific focus can to some extent be tailored to the person's particular skills and interests, and substantial assistance will be available to get 'up to speed'.

On the training/ testing side, we seek someone to work closely with Michael, Rob and myself to expedite our progress. This person needs to have a passion for cognitive science/psychology and be particularly motivated to work on our kind of developmental AGI. Depending on the level of experience, the candidate could work fairly autonomously, or alternatively, fill more of a supporting role.

We have a strong preference for the positions to be Los Angeles based, but we will consider remote candidates. Accommodation can be provided. Remuneration is primarily in the form of company equity, with only a subsistence level cash component. Working conditions are extremely flexible, and our environment allows for a lot of autonomy and job satisfaction.

At our current shoestring-level budget the project is funded for at least another 12 months.

Finally, we may be getting close to actually being able to use some remote, part-time help with specific aspects of our system (such as vision, sound, virtual world, etc.). Please subscribe to our tech list, if you are potentially interested in participating: http://groups.yahoo.com/group/a2i2Tech

Please contact me for further details - peter@optimal.org

Towards Increased Intelligence,

Peter Voss

January 1, 2003

AGI Project Review for 2002

Overview

Our current objective is to design, build and evaluate an initial proof-of-concept prototype that will demonstrate the feasibility of building AGI's based on my model of general intelligence. Work done during this past year has, in my opinion, served to confirm the validity of our general approach. While I am also more aware of the enormity of the task ahead, I feel even more confident now than before that no fundamental theoretical issues stand in the way of achieving high-level, autonomous artificial intelligence in the near-term.

In spite of significant start-up costs (mainly time), this first year showed excellent progress in creating our prototype's framework, infrastructure, and basic cognitive algorithms. We also gained a lot of valuable experience with our development environment. While individual component testing went well, integrated system testing/ training did not live up to our expectations. Reasonable progress was made on fleshing out the system design, and we managed to include many features that were not originally planned for this period.

Over the year we have had some excellent people work on the project and now have a solid, experienced team that can advance the project at an ever-increasing pace. However, the project would greatly benefit from additional managerial and (cognitive) test design capacity.

Financially we are in rather good shape, in that we can continue at our current (extremely low) burn-rate for at least another 8 to 12 months. While we could certainly benefit from substantially increased funding (primarily to be able to hire more staff), the pressure is not immediate, and - provided that we continue to make good progress - obtaining additional investments should become easier.

Our original milestone was to achieve 'Phase-1' functionality sometime early 2003. On the current development plan we are now working towards a greatly expanded series of milestones culminating in a 'Phase-2' objective, scheduled for around August 2003. Provided that we achieve our ambitious goals in this time-frame, we will be well positioned to take commercial advantage of our emerging AGI technology.

In the coming year we look forward to mastering the challenges ahead, and moving significantly closer to achieving Real AI.

Peter Voss

[ The full report is available to investors and other selected individuals ]

December 10, 2002

Friends of a2i2,

As mentioned in our last newsletter, we now have a powerful AGI (Artificial General Intelligence) test framework (including several dozen cognitive functions) and are shifting our focus to putting the system through its paces. Obtaining this practical feedback is invaluable for proving the soundness of our theoretical design, identifying limitations of our algorithms, and spawning new ideas. Our AGI testing/ training is now proceeding on three fronts: component level, virtual world systems, and real world interaction.

This brings me to the main purpose of this communication: We are now actively seeking one or two additional (Los Angeles based) team members to help spearhead our cognitive evaluation/ development program.

Main requirements are: very smart, innovative, self-motivated, and passionate about AGI/ theory-of-mind. This naturally means that they have read and thought about a fair amount of cognitive science - cognitive psychology, epistemology, AI, etc. Lots of practical experience may not necessarily be all that important because it is unlikely to be highly relevant to our particular AGI approach.

Candidates are expected to have read the description of our project ( http://adaptiveai.com/project/index.htm and http://adaptiveai.com/research/realai.htm ) and to have questions and comments based on these.

We are looking for people who find themselves in broad agreement with our project's particular focus, and who can fully commit to our specific approach. While we seek significant theoretical and design input to improve our system, we are not looking to explore fundamentally different models or theories at this time.

The two positions cover the full spectrum between bright, eager trainee AGI teacher and full-fledged, highly experienced senior partner in our venture. Both involve substantial hands-on implementation and experimentation. Programming skills are not a requirement, though they are somewhat desirable. Knowledge of animal or child psychometrics/ training would be a definite plus.

Note that the current level of our project implements low-level, low-performance proof-of-concept cognition, no higher than dog or baby level ability (at much, much lower motor-sensory resolution of course). Our focus during this phase of development is on adaptive perceptual, interactive learning, not abstract thinking, language or formal logic. Having said that, in order to aid trainer interaction and debugging (and for some other practical reasons), we do include some high-level, symbolic features in our design.

What we want is your passionate, dedicated contribution to our ambitious venture of creating Real AI.

What we offer is limited cash compensation (to cover minimal living expenses, if needed), significant equity involvement, a highly flexible self-directed work environment, and the chance to be part of the team poised to make considerable inroads into solving the core problems of General Artificial Intelligence - with all the potential rewards it entails.

Please contact me at peter@optimal.org if you are interested, or pass on this information to anyone suitable.

Till next time

Peter Voss

3-November-2002

Friends of a2i2,

Great news: A new investor! This will extend our (shoe-string budget) funding to at least middle of next year, probably longer.

Unfortunately, we were declined a NSF SBIR grant that took over two man-months to prepare for.

The only change to our team in the last two months, has been Shane Legg leaving to pursue his Ph.D. We wish him the best, and will miss his cheery manner!

Thanks to Louise Gold's efforts; our books, accounts, and other company legal requirements are now up-to-date, and we are in a better position to budget ahead.

With our overall project status now clearer, we are changing our focus:

With our original phase-1 objectives now nearer in sight, we are starting to develop detailed specs and to implement more advanced features and abilities -- we are heading for phase-2 and beyond!

Testing, training and tuning of current abilities is happening in parallel with development of enhanced cognitive features. A large part of our effort over the past 10 months was directed at creating our AGI development framework and associated tools. This investment is now paying off: we are now at the exciting stage where more and more of our time is spent of cognition, and less on programming.

In the light of these developments we are now looking for one or two additional team members who are particularly passionate about the cognitive aspects of this project: theory of mind and intelligence, epistemology, cognitive & developmental psychology (animal and human), etc. The central aspect of these positions is to help develop AGI tests and training programs, to run and evaluate these, and to help design and optimize cognitive structures and algorithms. We have already identified one potential candidate who may be joining us in LA soon.

In addition we are also open to discussing other partnerships, collaborations, and investments.

Peter Voss

8-Sept-2002

Friends of a2i2,

The ongoing pursuit of general intelligence in Los Angeles is going well!

The programmer's pow wow in late August proved to be highly productive in getting our multinational team in synch and up to date on the current code base and it's features. Both Shane Legg and David McFadzean were able to come from out of country to LA, Shane for several weeks beyond the meeting. During the week long pow-wow, many face-to-face meetings covering all aspects of the current system, the company and the road-map for the next 3-6 months were discussed. Also a much needed review of the last several months of development resulted in significant restructuring and unification of the code base. Now finished with those changes the team is tackling the difficult job of defining appropriate General Intelligence tests, to reveal strengths and weaknesses of the current system, and what development is most critically needed.

While there are still some major technical hurdles to clear, at present the