Meta to announce massive $23 billion move in race towards ‘superintelligence'

News.com.au

a day ago

Meta is preparing to drop a staggering $15 billion (A$23.09bn) on a stake in AI startup Scale AI, in a bold new play to push beyond current artificial intelligence capabilities and reach so-called 'superintelligence'.
The move would give Meta a 49 per cent stake in the company, which is currently led by 28-year-old Alexandr Wang.
The deal is yet to be officially confirmed, but multiple reports suggest Meta CEO Mark Zuckerberg is set to unveil the investment in the coming days.
Analysts say this is the behaviour of a 'wartime CEO', referring to the escalating technological arms race taking place between the world's most powerful companies and governments.
They all want the same thing, but nobody can tell exactly what happens once they get there.
Superintelligence refers to a hypothetical AI system that can outperform humans at all tasks, not just the specific functions currently delegated to today's large language models or image generators.
The industry isn't there yet, but given the gargantuan leaps we've witnessed over the past 18 months, it is getting increasingly likely we will see something major before the end of the decade.
Even as systems like GPT-4, Claude and Gemini dominate headlines, experts routinely point out their patchy reliability. Several language models still falter on complex reasoning and struggle with logic puzzles that your average Joe could solve.
At the moment, AI is about speed. It can effectively eliminate the legwork for a wide rage of everyday tasks performed on a computer. Humans are currently required to prompt their request, but there is still an issue of AI being sycophantic to the user. They are designed to impress, and therefore become confused at times when given a great deal of data, especially if some of it is conflicting.
At any rate, Zuckerberg wants in on the party.
Meta's attempt to leapfrog the crawl toward artificial general intelligence (AGI) is widely seen as an effort to re-establish dominance in an ecosystem now defined by competitors like OpenAI, Google DeepMind, and Anthropic.
It comes in the wake of Meta's ill-fated Metaverse experiment, which soaked up tens of billions in investment only to be largely shelved and mocked in equal measure.
Meanwhile, Scale AI recently made headlines for securing a deal with the US Department of Defense to develop ThunderForge, a military AI platform intended to support strategic planning in the Indo-Pacific and Europe. The company also counts Peter Thiel's Founders Fund among its early backers.
Observers say the mega-deal from Meta should reignite conversations in Europe about the need for publicly accountable AI research, something on par with CERN, the European particle physics laboratory.
Michael Wooldridge, Professor of the Foundations of AI at Oxford University, argued such an initiative would build trust through openness.
'There's a good argument that there should be a CERN for AI where governments collaborate to develop AI openly and robustly,' Prof Wooldridge said.
'That's not going to happen if it's developed behind closed doors. [AI] seems just as important as CERN and particle accelerators.'
Global arms race heats up, but oversight matters
With staggering sums of capital, military interest, and corporate strategy all converging, it is clear which way authorities want us to head as they scramble for AGI supremacy.
The race has been loosely compared to the frantic efforts in the 1940s to produce the world's first nuclear bomb.
Startups are being snapped up at record speed, university research labs are being drained of talent, and AI labs are increasingly moving into secrecy.
Global experts have already raised the alarm and called for robust oversight, but for those pessimistic about futurism, it has come as too little, too late.
In a report published ahead of the UN's highly anticipated 'Summit of the Future', pundits raised current lack of international oversight on AI.
Among the concerns are the very obvious opportunities for misuse, internal biases, and humanity's growing dependence.
One man known as the 'godfather of AI' famously quit Google in 2023 over concerns the company was not adequately assessing the risks, warning we could be walking into a 'nightmare'.
While the immediate benefits are already being seen in terms of productivity, the main concern is that we are charging full steam ahead towards an event horizon that is impossible to predict the outcome of.
What we do know is that those spearheading AI development are becoming absurdly wealthy incredibly quickly and thus hold more and more power over the trajectory of the planet as each day passes.
Around 40 experts, spanning technology, law, and data protection, were gathered by UN Secretary-General Antonio Guterres to tackle the existential issue head-on.
They say that AI's global, border-crossing nature makes governance a mess, and we're missing the tools needed to address the chaos.
The panel's report drops a sobering reminder, warning that if we wait until AI presents an undeniable threat, it could already be too late to mount a proper defence.
'There is, today, a global governance deficit with respect to AI,' the panel of experts warned in their report, stressing that the technology needs to 'serve humanity equitably and safely'.