Beyond Tomorrow: Exits From OpenAI, Startup Acquisitions & AI Robots—Weekly AI Roundup

Big Tech is looking to dodge regulators, again. That shouldn't come as any surprise to anybody. The U.S.' Federal Trade Commission and the European Union's regulators smell blood in the water, and it's a matter of time when they get there.

Advanced Micro Devices has been making a slew of acquisitions, with the latest being Europe's largest artificial intelligence lab, Silo AI. Apart from that, Accenture has been lapping up companies that work in the silicon design and engineering spaces.

Finally, are we going to get our very own Rosey The Robot from the Jetsons in the future? Fingers crossed! That's what recently-emerged-from-stealth startup Skild AI is planning for the world. The U.S.-based company is building scalable foundation AI models for robotics to be used across sectors.

Here is some AI news from the past week that's worth a read.

For a company that has invested $13.1 billion in OpenAI, the company behind ChatGPT, one would imagine that Microsoft would like a lot more say in the workings of OpenAI. Not really, as it turns out.

The Financial Times broke the news first (it's hard paywalled), confirming that both Microsoft and Apple were relinquishing their non-voting roles on OpenAI’s board. Going forward, the company will instead be holding regular meetings with major investors and partners, the company told the Washington Post in an email.

Why the sudden U-turn? Regulators are sniffing around, and they’re not happy. European regulators have previously said they intend to speak to rivals of Microsoft regarding its exclusive access to OpenAI’s technology. Even in the U.S., the Federal Trade Commission is opening an anti-trust investigation into Microsoft.

The Windows-maker is already facing a probe from the FTC for its failure to report the acquisition of Inflection AI, which is making machine learning and AI personalised.

Apple’s exit from the board role is pretty on brand. The company is deathly allergic to bad press and an inquiry into Microsoft puts the iPhone maker in the crosshairs as well.

It was probably a smart decision too, considering the company just settled an European Union antitrust lawsuit regarding the underlying technology to Apple Pay’s tap-and-go technology.

Back home, the FTC Chair Lina M Khan is no slouch either. She’s been very vocal in the past about bringing Big Tech to heel. Khan has been fairly consistent in opening investigations into large, industry-leading companies and has made it clear that she intends to follow through. TechCrunch has a conversation with Khan on their website, I recommend you checking it out if you want to understand how regulators are thinking about the intersection between Big Tech and AI.

As it stands, Nvidia is the clear winner in the race to provide AI chips. AMD is making sure it doesn’t fall behind.

The U.S. semiconductor company recently announced that it will acquire a Finland-based startup called Silo AI, for $665 million, according to a report from Reuters. According to the company’s website, Silo AI is “Europe’s largest private AI lab,” and partners with industries to build “smart devices, autonomous vehicles, industry 4.0, and smart cities.”

AMD’s acquisition of Silo AI’s is curious. Finland is part of the E.U. and would be subject to stringent technology regulation from the coalition. Experts both within the E.U. and outside have been saying that the region’s strict regulation will hamper innovation, compared to other regions.

As a result of the new law, companies are likely going to have to shift their strategies in the region and put in place coherent regulation within their own companies for fear of EU regulators coming down on them hard.

In other news, Accenture is also stepping up its AI game, with the acquisition of Cientra, a silicon design and engineering services company.

“Our acquisition of Cientra is our latest move to expand our silicon design and engineering capabilities and it underscores our commitment to helping our clients maximise value and reinvent themselves in this space,” said Karthik Narain, group chief executive—Technology at Accenture.

Check out NDTV Profit's Haripriya Sureban’s report here.

Everyone’s seen that meme about how they’d much prefer AI to take care of stuff like doing the laundry, cooking and daily chores that are really a tiring process. We can all agree that’s the dream. Luckily, there’s a company working on something just for that.

Skild AI, a robotics company building a scalable foundation model for robotics, has closed its Series A funding round with $300 million at a valuation of $1.5 billion, according to Businesswire.

Skild is apparently using 1,000 more data points to train its foundation model compared to its competitors. They say that their model isn’t going to be hyper specialized to specific requirements but will serve as a "general-purpose" brain. As a result, their technology will enable the “use of low-cost robots across a broad range of industries and applications.” Guess we’re not too far from coming home to see our laundry dried and pressed, dinner on the table and the bed made.

But seriously though, a robot butler in our lifetimes?

We’re still probably decades away from seeing a robot Man Friday, but one can dream. We’ve yet to figure out issues with AI, including understanding concepts and applying them to new situations. After all, nearly all the AI that exists in the world currently is still based on pattern recognition. So how do we move from that and get ahead.

Turns out, researchers figured out a potential way to teach an AI to learn concepts last year. It’s called Meta-learning for Compositionality. Two researchers, one from New York University and the other from Spain’s Pompeu Fabra University, found you can basically teach “compositional generalisation”. In simpler words, that involves taking interconnecting concepts and applying them in new settings. To do this, you’ve basically got to sit a neural network (the engine on which stuff like ChatGPT works) down, and make it practice stuff. That’s also kinda how machine learning works in general.