OpenAI's New System for Determining How Close AGI Is

OpenAI has introduced a new five-level system to benchmark progress towards achieving Artificial General Intelligence (AGI), with current AI at level one (chatbots) and on the verge of level two (reasoners with human-level problem-solving). The system aims to provide insight into AI development stages and associated risks. Additionally, Microsoft and Apple are withdrawing from their OpenAI board observer roles due to antitrust concerns, reflecting increased regulatory scrutiny in the US and Europe. This move suggests these companies will seek alternative methods for maintaining insight into OpenAI's developments without exacerbating regulatory pressures.

• OpenAI has introduced a new leveling system to help understand the progress towards achieving Artificial General Intelligence (AGI).
• The system includes five levels that benchmark the stages of AI development, providing insights into the risks and capabilities at each stage.
• This initiative follows recent organizational changes at OpenAI, including the firing and rehiring of Sam Altman and the departure of key team members.

• The leveling system was introduced at an all-hands meeting and confirmed by an OpenAI spokesperson.
• It aims to provide a clearer understanding of how close current AI technologies are to achieving AGI.

"OpenAI has developed a new system to explain where we are in progressing towards AI that achieves artificial general intelligence or AGI."

• This leveling system is designed to benchmark the progress toward AGI, highlighting the stages and associated risks.

• Level one represents the current stage of AI, characterized by chatbots that can use conversational language.
• This level has been the standard for nearly two years, especially post the release of ChatGPT.

"The stages of artificial intelligence start where we are now at level one with chatbots. These are AIs that can use conversational language."

• The current level (level one) is marked by the use of conversational AI, indicating limited risks compared to higher levels.

• OpenAI's expressed mandate is to produce safe and beneficial AGI for all.
• The new leveling system, while not comprehensive, provides insight into how OpenAI perceives risks at various stages.

"It is of course the expressed mandate and purpose of OpenAI to produce safe and beneficial AGI for all."

• This mandate underpins the development of the leveling system, aiming to ensure safety and benefits in AI advancements.

• The introduction of the leveling system follows significant organizational changes at OpenAI.
• These changes include the firing and rehiring of CEO Sam Altman and the departure of key members from the super alignment team.

"Ever since the crisis last fall with Sam Altman being fired and then rehired and more recently with Ilia leaving the company alongside basically the entire super alignment team."

• These events have influenced OpenAI's approach to long-term AI alignment and safety, leading to the development of the leveling system.

• The new system provides a framework for understanding the risks associated with different stages of AI development.
• It offers a preliminary insight into how OpenAI views the progression towards AGI and the associated safety considerations.

"This new leveling system certainly isn't a comprehensive approach to that but does give a little bit more insight into how the company sees risks at various stages."

• The system is not exhaustive but serves as a tool for gauging progress and risks in AI development.

• OpenAI's new leveling system is a significant step in mapping the journey towards AGI.
• It reflects the company's commitment to transparency and safety in AI development amidst recent organizational upheavals.

"Fundamentally what OpenAI is saying with these levels is that for almost two years now this is the level that we've been at level one."

• The leveling system highlights the current state of AI and sets the stage for understanding future advancements and risks.

• GPT-4 and GPT-3.5 are identified as having significant capabilities and potential risks.
• The discussion around these models focuses on their problem-solving abilities and potential future developments.

"This sort of GPT 4.0, GPT 3.5 type of capacity is a real risk but what comes next."

• Highlights the perceived risks associated with the advanced capabilities of GPT-4 and GPT-3.5.

• OpenAI has introduced a leveling system to classify the stages of AI development.
• The system includes five levels, each representing a different stage of AI capabilities.

"The level two stage on OpenAI system is called reasoners with human-level problem-solving."

• Level two is defined by systems that can solve basic problems as well as a human with a doctorate level education.

• Level two AI systems are expected to have human-level problem-solving skills without access to external tools.
• OpenAI demonstrated a research project involving its GPT-4 AI model, showcasing human-like reasoning skills.

"Bloomberg writes this refers to systems that can do basic problem-solving tasks as well as a human with a doctorate level education who doesn't have access to any tools."

• Level two AI systems are compared to humans with advanced education but without external aids.

"Company leadership gave a demonstration of a research project involving its GPT-4 AI model that OpenAI thinks shows some skills that rise to human-like reasoning."

• OpenAI's demonstration indicates that GPT-4 may possess reasoning abilities comparable to humans.

• Level three AI systems, referred to as "Agents," can take autonomous actions.
• This level has been a major focus of excitement and anticipation.

"You may note that level two is not the thing that everyone has been excited about for more than a year which is agents that comes at level three."

• Emphasizes that the AI community is particularly excited about the potential of level three systems.

• Level four AI systems, known as "Innovators," can assist in the invention process.
• These systems are expected to contribute to creative and novel solutions.

"At level four we have innovators AI that can aid in invention."

• Innovators are AI systems that can help in creating new inventions and solutions.

• Level five AI systems, termed "Organizations," can perform the work of an entire organization.
• This represents the highest level of AI capability in OpenAI's leveling system.

"At level five organizations AI that can do the work of an entire organization."

• Level five signifies AI systems that can manage and execute tasks typically handled by an entire organization.

• The timeline for reaching each level of AI development is not clearly defined.
• OpenAI is continuously testing new capabilities internally.

"This is not necessarily a final determination this was created by a set of OpenAI Executives and is meant to be a work in progress."

• The leveling system is a provisional framework created by OpenAI executives and may evolve over time.

"It doesn't appear at least that we've seen any sort of discussion around how long OpenAI believes it will take to get to each of these different stages."

• There is no specific timeline for achieving the different stages of AI development.

• There are claims and speculations about interactions with GPT-5, though these are not confirmed.

"I did see on Twitter today someone who claimed that they had sources in their DM saying that they had actually interacted with GPT-5."

• Unverified claims about GPT-5 interactions suggest ongoing developments and interest in future AI models.

• Metaculus, a forecasting platform, predicts OpenAI will announce GPT-5 on January 13, 2025.
• The prediction is based on the consensus of 652 forecasters.

"The current betting among 652 forecasters around when OpenAI will announce GPT-5 is January 13th of next year, 2025."

• The quote highlights the specific date predicted for the GPT-5 announcement and the number of forecasters involved.

• Discussion on AI capability levels, ranging from level 2 to level 5.
• Debate on whether achieving one level automatically implies achieving subsequent levels.
• Different opinions on the definitions and implications of these levels.

"Mario Kistra writes, 'Strange list. It seems to me that as soon as you get to level two, you also get 3, 4, and 5 unless there is a weird definition of human level. Humans can do all these things, so human level AI should be able to do all those things.'"

• Mario Kistra suggests that achieving level 2 should inherently lead to achieving levels 3, 4, and 5, questioning the definitions of human-level AI.

"Matt Garcia points out that level five organizations kind of sounds like Advanced super intelligence."

• Matt Garcia associates level 5 organizations with advanced superintelligence, indicating a higher degree of AI capability.

"Pseudonym writes, 'This would be a better way to structure an atome agent flow.'"

• Pseudonym discusses structuring AI agent flows, implying a need for better organization.

"Steph writes, 'Once you reach level two, level three should be a piece of cake. I'm probably wrong, but just by making AI have access to APIs, it should become an agent if it can already plan things.'"

• Steph believes that once level 2 is achieved, level 3 should be easier, especially with API access enabling planning capabilities.

• Microsoft and Apple are removing their board observer roles at OpenAI.
• Historical context: Microsoft had a longstanding relationship with OpenAI and took a formal board observer role after a debacle in November.

"Microsoft and Apple are both getting rid of their OpenAI board observer roles."

• The quote indicates a significant change in the relationship dynamics between OpenAI and these tech giants.

"Microsoft has obviously had a long-standing relationship with OpenAI, but it wasn't until after the debacle last November that it took a formal board observer role."

• Provides background on Microsoft's involvement with OpenAI, emphasizing the formalization of their role post-debacle.

• Anticipation of further discussions on AI capability levels.
• Uncertainty about whether these discussions will gain traction or remain theoretical.

"I will be interested to see whether anyone else picks up this discussion over the next couple of weeks or whether this will mostly remain on the drawing board."

• The quote expresses curiosity about the future discourse on AI capability levels and their practical implications.

• Antitrust scrutiny has influenced the relationship between big tech companies and AI collaborations:
  • Apple's potential board Observer role with ChatGPT on iPhone was reconsidered due to antitrust concerns.
  • Regulatory scrutiny has been increasing, particularly in the US and Europe.

"Antitrust scrutiny has been on the rise in general."

• This quote highlights the growing concern and regulatory focus on antitrust issues within the tech industry.

"Microsoft's board Observer role seems to have intensified regulatory scrutiny of its open AI relationship both in Europe and in the US."

• This quote indicates that Microsoft's involvement with OpenAI has attracted significant regulatory attention, which might have influenced the decision to avoid similar roles for Apple.

• Board Observer roles and their implications:

  • These roles do not have the power to influence decisions or participate in voting but provide visibility and insight.
  • The increased regulatory pressure associated with such roles may not be worth the hassle for the companies involved.

"A board Observer role is exactly that, it's an observer role. It doesn't have any power to actually shape or influence things or be a part of any sort of voting."

• This quote clarifies the limited scope and influence of a board Observer role, emphasizing that it is primarily about gaining insight rather than control.

• Regulatory bodies and their influence:

  • The European Commission's antitrust leader concluded that Microsoft hadn't acquired control of OpenAI through its board Observer role.
  • If regulatory bodies triggered the shakeup, it was likely the FTC in the US or the Competition and Markets Authority in the UK.

"The European commission's antitrust leader has said that her team had concluded that Microsoft hadn't acquired control of open AI after getting this board Observer role."

• This quote underscores the decision by the European Commission, suggesting that the regulatory concerns might have originated from other bodies like the FTC or the UK's Competition and Markets Authority.

• Current state of regulatory discourse:

  • While regulatory discussions around AI have slowed in the US, antitrust conversations have intensified.

"The regulatory discourse around AI has certainly slowed down in the US this year, the anti-rust conversation has done nothing but heat up."

• This quote contrasts the slowdown in AI-specific regulatory conversations with the increasing intensity of antitrust discussions, indicating a shift in focus among regulators.

• Summary of the episode:
  • The episode concludes with a brief acknowledgment and appreciation for the listeners, indicating that the discussion was meant to provide quick insights into the current state of regulatory concerns affecting tech companies and AI collaborations.

"For now though that is going to do it for this quick episode today appreciate you listening as always and until next time peace."

• This quote wraps up the episode, thanking the audience and signaling the end of the discussion on the topic.