This time it’s starting to seem real. In a recent interview with YC chief Gary Tan, OpenAI Sam Altman, CEO common his ideas on AGI (artificial general intelligence), suggesting that AGI could be within reach as early as 2025.

“I think we’ll get there faster than expected,” he said, highlighting OpenAI’s accelerated progress.

“We actually know what to do…it will take time, it will be difficult, but it is extremely exciting,” Altman said, reflecting on their recent advancessignaling that internal developments have already exceeded public expectations.

OpenAI appears to have cracked AGI internally

OpenAI’s strategic directionrooted in scaling laws and deep conviction, played a vital role in its progression towards AGI. He highlights the underestimated value of “a fairly extreme level of conviction on a bet.”

“We said from the beginning that we were going to take on AGI at a time when, on the ground, you weren’t allowed to say that because it seemed incredibly crazy” recalled Altman, pushing the boundaries of research.

Furthermore, he said OpenAI had fewer resources than deep mind and others. “So we said okay, they’re going to try a lot of things and we just need to pick one and really focus,” he said. added.

“I’ve heard people claim that Sam is just creating hype, but from what I’ve seen, everything he says fits the middle view of OpenAI researchers on the ground,” said Noam Brownresearcher at OpenAI.

Even the CFO of OpenAI, Sarah Brotheragrees. “One of the best meetings I get to attend is our research meetings, and it would blow your mind to see what’s already coming,” she said, discussing the topic . abilities of o1 And GPT models to come.

“We have the plan in place. I think if Sam were sitting in this seat he would tell you that (AGI) is closer than most people think,” Friar added in a recent interview with Bloomberg.

If you still don’t believe it, here’s a viral video from OpenAI from five years ago, where we see agents playing hide and seek.

THE paper published alongside explored a framework in which competition in simple contexts can allow AI to develop sophisticated skills, potentially leading to human or animal intelligence.

In the paperagents employ self-supervised learning with multiple sets of strategies emerging via hide-and-seek and multi-agent competition. Multi-agent competition scales well to complex environments, leading to skills closer to human tasks than other reinforcement learning methods.

So, five years ago, OpenAI reached the third stage (Agents). “This move from one (chatbots) to two (reasoners) took a while, but I think one of the most exciting things about two is that it allows you to enable level three relatively quickly and the agent experiences that I think we hope this technology will eventually enable. , will have a considerable impact”, said Altman, hinting at an AGI or ASI future.

The latest OpenAI model, o1marks an important step towards AGIAltman expressing newfound confidence in achieving human-level reasoning and progressing to the next phase of AGI, “Level 3 = Agents”, as he stated: “We have reached human-level reasoning and will now move to level 3.”

During the recent AMA session on Reddit, Altman said AGI is achievable with current hardware. A few days ago, OpenAI announced plans to launch its first in-house AI chip by 2026, partnership with Broadcom and TSMCas part of a strategic move to diversify its reliance on NVIDIA’s dominant GPUs and optimize its AI infrastructure amid rising costs and supply constraints.

Last month, NVIDIA book its advanced Blackwell AI chips to OpenAI and Microsoftmarking a crucial step in accelerating AGI development with faster training and superior inference performance.

AGI drama unfolds

Currently, OpenAI’s contract with Microsoft includes a crucial clause: whether OpenAI achieves artificial general intelligence (AGI) – defined as a machine corresponding to the capabilities of the human brain – Microsoft loses access to OpenAI technologies.

Originally intended to prevent potential misuse by Microsoft, OpenAI executives now see it as leverage for a better deal. “Depending on the terms of the contract, the OpenAI board of directors could decide on the arrival date of AGI,” adding a strategic dimension to their partnership.

“Microsoft should never have agreed to such a stupid clause. What will stop the OAI Board of Directors from calling on AGI to terminate the contract? said Gary Marcuscriticizing this arrangement.

Co-founder of IFP Caleb Watney echoes similar sentimentsobserving that “OpenAI threatens to trigger its famous ‘AGI Achieved’ vulnerability, mainly to get out of the Microsoft contract and have leverage to renegotiate calculation prices.

At the same time, Microsoft is very unlikely to stop OpenAI from realizing its AGI dreams.

Already AGI: It feels like yesterday again when Altman shared the stage with Microsoft chief Satya Nadella at his first-ever developers conference – Development Day 2023where he says:

“I think we have the best technology partnership. I’m excited that we can build AGI together,” Altman said.

Let’s get back to reality: OpenAI would barely have survived without Microsoft. Much of the training computing today is handled by Microsoft, which provides Azure servers for training. OpenAI models.

In 2022, when OpenAI launched ChatGPT—who will soon complete two years–Altman credited Microsoft for this. “Microsoft, and particularly Azure, doesn’t get enough credit for OpenAI launches. They do an incredible job of achieving this; we are deeply grateful for this partnership. They have built by far the best AI infrastructure there is,” said Altman.

But OpenAI and Microsoft are not alone

Google now overtakes OpenAI in the race for AI advancements, with rapid deployments like Gemini 1.5, Gem 3, CarnetLMand the DataGemma model that addresses hallucinations. Their most anticipated release to date, Gemini 2is described by Logan Kilpatrick as having “better quality of reasoning and a longer pop-up window.” This approach is also supported by Sundar Pichai reports on 14x increase in the use of the Gemini API, it shows that the race for artificial intelligence is more competitive than ever.

In the meantime Meta accelerates its AI developments with the intention of publishing Lama 4 by early 2025showcasing advances in memory, contextual capabilities, and multimodality. This corresponds to its ambitious desire to Autonomous Machine Intelligence (AMI).

“Who doesn’t want to be the first on this mountain? » said NVIDIA chief Jensen Huang, in a recent interview with No Priors, said that the the race to reach AGI becomes fiercewith major players like – OpenAI, AnthropicxAI, alongside Google, Meta and Microsoft, are all vying for position.

“The price for completely reinventing intelligence… is too great not to attempt,” he added, emphasizing that scaling laws and massive computing advances are crucial.

“We are close to artificial general intelligence… but even if we could debate whether it really is general intelligence, just getting close to it would be a miracle… Anything goes be difficult…but nothing is impossible,” Huang concluded.