Searching for the next breakthrough in AI

TORONTO — For several days, AI chip giant Nvidia has sat on its throne as the world’s biggest company, but behind its stunning success are questions about whether new entrants can lay claim to the AI ​​bonanza.

Nvidia, which makes processors as the only option for training large language models for generative artificial intelligence, is now the newest member of Big Tech, and its stock market debut has lifted the entire sector.

Even the second tier of Wall Street’s tech industry has fallen on par with Nvidia, with Oracle, Broadcom, HP and a host of other companies seeing their stock valuations skyrocket despite sometimes shaky earnings.

Amid the pouring of champagne, startups seeking the attention of Silicon Valley venture capitalists are being asked to innovate – but with no clear indication of where the next chapter on artificial intelligence will be written.

When it comes to generative AI, there are still questions about what exactly will be left for companies that are not existing modelers – a field dominated by Microsoft-backed OpenAI, Google and Anthropic.

Most agree that competing with them head-on might be a fool’s errand.

“I don’t think there’s a great opportunity right now to start an entry-level AI company,” Mike Myer, founder and CEO of technology company Quiq, said at the Collision technology conference in Toronto.

Some have tried to create applications that leverage or imitate the capabilities of existing large models, but Silicon Valley’s biggest players have fought back.

“What’s troubling is that people don’t differentiate between apps that are model killers as you increase their capabilities and those that really add value and will be around in 10 years,” said veteran venture capitalist Vinod Khosla .

“I can’t keep up”

Tough talker Khosla is one of OpenAI’s early investors.

“Grammar can’t keep up,” Khosla predicted about spelling and grammar checking apps and others like them.

Those companies that only “thinly wrap” what AI models can offer are doomed, he said.

Khosla said one field ripe for entry is chip design, as artificial intelligence requires increasingly specialized processors that provide very specific powers.

“If you look at the history of chips, we’ve really focused mainly on more general chips,” Rebecca Parsons, chief technology officer at consultancy Thinkworks, told AFP.

Providing more specialized processing for multiple AI requirements is an opportunity seized by Groq, a hot start-up that has built chips to implement AI rather than train it, or infer it – the specialty of Nvidia’s world-dominant GPUs.

Groq CEO Jonathan Ross told AFP that Nvidia won’t be the best at everything, even if it has no competition in generative AI training.

“Nvidia and (its CEO) Jensen Huang are like Michael Jordan… the best in the history of basketball. But inference is baseball, and we try to forget about the time when Michael Jordan tried to play baseball but wasn’t very good at it,” he said.

The next opportunity will be highly specialized artificial intelligence, which will provide knowledge and know-how based on proprietary data that will not be hijacked by voracious big tech.

“Open AI and Google won’t build a construction engineer. They are not going to create products like a primary care doctor or a mental health therapist,” Khosla said.

Profiting from highly specialized data is the basis of Cohere, another of Silicon Valley’s hottest startups, which offers specially developed models to companies that fear that artificial intelligence will get out of control.

“Enterprises are skeptical of technology and averse to taking risks, so we need to earn their trust and prove to them that there is a way to adopt this technology that is reliable, trustworthy and safe,” Cohere CEO Aidan Gomez told AFP. .

When he was just 20 years old and working at Google, Gomez co-authored the groundbreaking paper “Attention Is All You Need,” which introduced Transformer, the architecture behind popular large-language models like OpenAI’s GPT-4.

The company has received funding from Nvidia and Salesforce Ventures and is valued at billions of dollars.