It's an important step forward, but it doesn't make a leap in the finish line. That's how Openai CEO Sam Altman explained his latest upgrade to ChatGpt this week.
The race Altman mentioned was artificial general information (AGI), by Openai's definition, which is the theoretical state of AI in which highly autonomous systems can do human work.
By describing the new GPT-5 model that powers the Power ChatGpt as a “critical step on the road to AGI,” he nevertheless added a hefty warning.
“[It is] It lacks something very important, very important,” says Altman, who says that while these systems are impressive, such as the inability to “learn continually” the models even after launch, they still haven't solved the autonomy that allows them to work full-time.
Openai's competitors also lavish the same targets and wash them off in billions of dollars, putting a strain on the tape. Last month, Facebook Parent Meta CEO Mark Zuckerberg said the development of Superintelligence, another theoretical state of AI where systems go far beyond human cognitive abilities, is “now visible.”
On Tuesday, Google's AI unit outlined the next steps to AGI by unveiling an unpublished model that trains AIS to interact with compelling real-world simulations, but humanity, with significant advances in humanity, announced an upgrade to the Claude Opus 4 model.
So, where does this leave the race in AGI and Super Intelligence?
Technology analyst Benedict Evans says the competition towards the theoretical state of AI is happening against the background of scientific uncertainty.
Describing AGI as “as much as thought experiment as it is technology,” he says:
He adds: “We're building an Apollo program, and we don't really know how gravity works, how far the moon is, how far the rockets are, but if we keep making the rockets bigger, we can keep getting there.”
“To use the present moment, it's very vibe-based. All of these AI scientists really tell us what their personal vibe is about whether to reach this theoretical state, but they don't know. That's something that even wise experts say.”
However, Aaron Rosenberg, a partner at venture capital firm Radical Ventures, said its investment includes the major AI companies' cooperation – and the previous head of strategy and operations at Google's AI unit DeepMind, could achieve a more limited definition of AGI at the end of the decade.
“If we define AGI more narrowly as human-level performance at at least 80% of economically relevant digital tasks, I think it's within reach over the next five years,” he says.
Matt Murphy, partner at VC Firm Menlo Ventures, says the definition of AGI is a “moving target.”
He adds: “I think the race will continue to unfold for years to come.
Even without AGI, the generated AI system in circulation makes money. The New York Times reported this month that Openai's annual recurring revenue reached $13 billion (£1 billion), up from $10 billion in the summer, and could pass $200 billion by the end of the year. Meanwhile, Openai is reportedly discussing the sale of shares held by current and former employees, which value it at around $500 million, exceeding Elon Musk's SpaceX price.
Some experts have seen statements about tight systems that appear to generate unrealistic expectations, but are distracted from more pressing concerns, such as ensuring that systems currently deployed are reliable, transparent and unbiased.
“The Rush, which advocates for “close” among major tech companies, reflects more about competitive positioning than actual technical breakthroughs,” says David Budder, director of the Institute of Data Science at New Jersey Institute of Technology.
After the newsletter promotion
“We need to distinguish between real advancements and marketing narratives designed to attract talent and investment. From a technical perspective, we see impressive improvements in certain abilities: better reasoning, more refined planning, and enhanced multimodal understanding.
“But well-defined superintelligence represents a system that exceeds human performance across almost every cognitive domain. We are not near that threshold.”
Nevertheless, major US tech companies continue to try and build systems that fit or exceed human intelligence for most tasks. According to the Wall Street Journal, Google's parent alphabet, Meta, Microsoft and Amazon alone will spend nearly $400 million on AI this year.
Rosenberg admits he is a former Google Deepmind employee, but says there are great benefits to the data, hardware, infrastructure and a variety of products that hone the technology, from search to maps to YouTube. However, the advantages are small.
“In frontiers, as soon as innovation occurs, everyone else adopts it right away. It's hard to get a big gap right now,” he says.
It is also a global race, or rather a contest, including China. Deepseek has come from anywhere this year, unveiling the Deepseek R1 model and boasts “strong and interesting reasoning behavior” that rivals Openai's best work.
Large companies looking to integrate AI into their businesses are taking notes. Saudi Arabia, the world's largest oil company, uses Deepseek's AI technology at its major Data Centre, saying it is “making a real difference” in its IT systems, making the company more efficient.
According to artificial analysis, six of the top 20 leaderboards (ranking models according to their range of metrics, such as intelligence, price, speed, etc.) are Chinese. Six models are developed by Deepseek, Zhipu AI, Alibaba and Minimax. On the video generation model leaderboard, six of the top 10 (including seedlance from current leader, Baitedan) are also in Chinese.
Microsoft's president Brad Smith, who bans Deepseek, told a US Senator's hearing in May that adopting the AI model globally is a key factor in determining which countries win the AI race.
“The number one factor that defines whether the US wins this race or not is its technology being most widely adopted in other parts of the world,” he said, adding that the lessons from Huawei and 5G are “hard to replace” for anyone establishing leadership in the market.
That means that debate aside the feasibility of close systems, a massive amount of money and talent is poured into this tribe of the world's two biggest economies, and tech companies continue to run.
“Looking back at 2020 five years ago, it was a blasphemous thing to say that Agi is on the horizon. It's been crazy to say that. It seems like more and more consensus to say that we are on that path now.”
