- OpenAI and Meta are preparing to unveil new versions of their AI models, claiming advancements in reasoning and planning, vital for achieving superhuman cognition in machines.
- Both companies are gearing up to release their updated large language models, with Meta launching Llama 3 and OpenAI expecting to introduce GPT-5 soon.
- The progression towards reasoning and planning capabilities in AI models signifies a crucial step toward achieving Artificial General Intelligence (AGI), enabling machines to handle more complex tasks and operate at a higher level of intelligence.
OpenAI and Meta are poised to release their respective new versions of artificial intelligence (AI) models, claiming these models will be capable of reasoning and planning, key steps toward achieving superhuman cognition in machines.
Executives from OpenAI and Meta announced this week that both companies are preparing to roll out the next iterations of their large language models, which power generative AI applications.
Also read: Sam Altman quits OpenAI venture fund under new structure
Also read: Microsoft, OpenAI plot US$100 billion Stargate AI supercomputer
Meta will begin rolling out Llama 3 in the coming weeks
Meta stated that it will begin rolling out Llama 3 in the coming weeks, while OpenAI indicated that its next model, expected to be called GPT-5, will be released “soon.”
Meta’s Vice President of AI Research, Joelle Pineau, said, “We are working on figuring out how to make these models not only talk but really reason and plan… have memory.”
The next generation of GPT will make progress in reasoning
Brad Lightcap, COO of OpenAI, stated in an interview that the next generation of GPT will make progress in solving “hard problems” like reasoning, “We will start to see AI able to handle more complex tasks in more sophisticated ways, I think we’re just scratching the surface of these models’ reasoning abilities.”
Lightcap added that current AI systems are “very good at one-shot small tasks,” but their capabilities are still “fairly limited.”
As tech companies ramp up their efforts to advance the technology, progress is accelerating, with competition to create more complex generative AI models – capable of producing human-like text, images, code, and videos – intensifying. This year, companies like Google, Anthropic, and Cohere have released new waves of large language models.
Reasoning and planning capabilities are crucial steps in AGI
Reasoning and planning capabilities are crucial steps toward achieving “Artificial General Intelligence (AGI).” AGI, possessing human-level cognition, enables chatbots and virtual assistants to complete sequences of relevant tasks and predict the consequences of their actions.
Yann LeCun, Meta’s Chief AI Scientist, stated at an event in London on Tuesday, April 9, that current AI systems “generate words one after another without thinking or planning,” finding it challenging to handle complex problems or retain information over long periods, thus still “making stupid mistakes.”
He mentioned that enhancing reasoning capabilities means AI models will “search for possible answers,” “plan sequences of actions,” and build a “mental model of what the consequences of its actions might be.” LeCun added that these abilities are a significant missing piece in current models, and they will take machines to the next level of intelligence.
Lightcap of OpenAI stated that there will soon be more to say about the next version of GPT, “I think over time we’ll see these models moving toward longer, more complex tasks, which requires them to have better reasoning capabilities.”
