Cloud Hosting

This article is from the free weekly Barrons Tech email newsletter. Sign up here to get it delivered directly to your inbox.

Smarter Siri. Hi everyone. Excitement over generative artificial intelligence has surged since OpenAIs release of ChatGPT. But I often wonder if AI is improving dramatically, why are our voice recognition assistants still so dumb?

Most people dont use Amazons (ticker: AMZN) Alexa or Apples (AAPL) Siri for anything beyond simple tasks like setting timers or getting a weather update. The two services rarely offer quality responses to factual queries.

But now there may be an app for that. In May, start-up Inflection AI released its Pi chatbot for iPhone. The name stands for personal intelligence. Its trying to become a more thoughtful and supportive AI companion. Users can interact with Pi through voice or text.

I recently started using Pi and was struck by how much better it is for fact-based questions than Siri. You can choose between six different AI voices, and it does a good job of carrying on a natural back-and-forth conversation. Pis answers are concise and informative. And the verbal interface seems like a natural evolution for large-language-model AI technology, compared to ChatGPTs text-based format. It is an experience that feels right out of a Star Trek-like future.

Advertisement - Scroll to Continue

Inflection AI has an impressive pedigree. The founders came from DeepMind, an AI company renowned for its expertise, which was acquired by Alphabet s (GOOGL) Google. Other employees previously worked at OpenAI and Meta Platforms (META). The company has top notch investors too. In June, Inflection AI announced it had raised a $1.3 billion funding round from Microsoft (MSFT), Nvidia (NVDA), Reid Hoffman, Bill Gates, and Eric Schmidt.

Barrons Tech recently spoke to Inflection AI co-founder and CEO Mustafa Suleyman. We discussed the state of AI technology, the competitive landscape for AI semiconductors, his recent meeting with President Joe Biden on AI regulation, and his optimistic vision for the future of AI.

[Inflection AI was among the seven leading AI companiesincluding Amazon , Meta Platforms, Microsoft, Google, and privately held firms OpenAI and Anthropicthat recently made voluntary commitments to responsibly develop AI.]

Advertisement - Scroll to Continue

Suleyman is a venture partner at Greylock Partners. Earlier in his career, he co-founded DeepMind. At Google, he was vice president of AI products & AI policy. His upcoming book The Coming Wave, about AIs societal ramifications, is scheduled for release on Sept. 5.

Here are highlights from our conversation with Suleyman:

Barrons: How was the AI summit at the White House last week?

Advertisement - Scroll to Continue

Suleyman: There was a great energy from the administration and from the companies. I think on both sides there was a sense this is unprecedented times. The pace of evolution is incredible. That justifies optimism and excitement, but at the same time its a moment for the precautionary principle.

What were the big takeaways?

The voluntary commitment [to test] these models [with internal and external teams] is a pretty significant development. If you think about the way it works in security today, most of the time when vulnerabilities are disclosed there is a 60 day window to share the exploits among the companies. Thats a cultural practice. There is no regulation around that. Its a norm.

Advertisement - Scroll to Continue

You can imagine a similar kind of culture evolving here [with AI companies] where frontier model development companies identify weaknesses in their model and they share those exploits with the other companies.

Thats a huge positive contribution to improving the probability of good outcomes. You can think about that happening in the context of bioweapons development, cyberweapons development, and nuclear and chemical weapons development. Theres plenty of types of conversation that we would not want to make it really easy for a bunch of people to suddenly get educated on how to do these harmful things.

Voice conversation seems to be the natural evolution for AI chatbots like Inflection AIs Pi. Whats your vision for Pi going forward?

Advertisement - Scroll to Continue

I think in the future everyone is going to have their own personal AI. Its going to have three types of capability. Its going to have an EQa personality of being kind and compassionate. Its going to have IQ. It will have knowledge with deep expertise in many different areas. And in time it will have AQ, an action quotient. Its going to be able to take actions on your behalf. Book things, buy things, plan or organize.

Thats going to be where these things end up with those three capabilities. And Pi, we think, is going to be one of the first to get to that user experience.

What kinds of innovations do you see in large-language model AI technology over the next couple years?

We are about to train models that are 10 times larger than the cutting edge GPT-4 and then 100 times larger than GPT-4. Thats what things look like over the next 18 months.

Thats going to be absolutely staggering. Its going to be eye-wateringly different. I think we can largely eliminate hallucinations [chatbots tendency to make up facts and information] in the next two years.

Nvidia is one of the financial backers of your company. So let me ask you: What is the state of the AI chip market and how defensible is Nvidias leadership position?

The GPU [graphics processing unit] market is dominated by Nvidia . Thats because they produce the best chipshands downby a long way. They also provide the best software in CUDA, which is Nvidias development software that sits on top of the chip.

Nvidia is in a very strong position. They have a lot of experience developing these chips after many generations. It takes a long time to de-bug these chips. We have the largest cluster in the world of Nvidia H100 chips today. That means we can train the biggest, the best, and fastest models.

AMD (AMD) is promising for inference, but not so much for training. Inference means serving the model. They have a larger memory on those chips, so you can serve bigger models. AMD hasnt figured out how to daisy chain the chips together well for big training jobs. Nvidia has thought about parallelization and [high bandwidth connection] networking [better].

What are your predictions on how AI can lead to prosperity? But also, on the flip side, how can it disrupt society?

I do think its going to be the most productive decade in the history of our species. Anyone who is a creator or an inventor is now going to have a compadre who gets their domain.

People who are trying to be productive are now going to have an aide that is going to turbocharge their productivity. Thats going to save people an insane amount of time. Its going to make us much more creative and inventive.

On the flip side, anyone who has an agenda to cause disruption, cause chaos, or spread misinformation, is also going to have the barriers of entry for their destabilization efforts lowered.

Technology tends to accelerate offense and defense at the same time. A knife can be used to cut tomatoes or to hurt somebody. Thats the challenge of the coming wave. Its about containment. How do nation states control the proliferation of very powerful technologies, which can ultimately be a threat to the existence of the nation state if they are left unchecked?

Thanks for your time, Mustafa.

Write to Tae Kim at tae.kim@barrons.com or follow him on Twitter at @firstadopter.

Continued here:
AI Could Spark the Most Productive Decade Ever. Nvidia Could Benefit. - Barron's