Technology

➚ MOVE FAST: Plot line. Sam Altman’s chip ambitions are coming into focus, Reuters reports. Rather than try to build its own AI chips, OpenAI is working with TSMC to build inference products, similar to Google’s custom Tensor processors.

➘ BREAK THINGS: Plot thickens. Elon Musk’s PAC has been secretly sending text messages and posting Facebook ads that claim to come from the Harris campaign, 404 Media revealed. The ads promote policies that Harris doesn’t support, such as mandatory gun buybacks.

A recent report from the Center for Policy Studies showed that chat bots from OpenAI, Anthropic and Meta all lean significantly to the political left. This debate has been going on since ChatGPT launched.

Often, we look at these chatbots through the lens of the web and social media. But AI is fundamentally different. It’s not going to be defined by winner-take-all platforms that lock consumers into a small handful of walled gardens.

The way to view these models is that they’re reasoning engines you can fine-tune. So if you want a conservative chatbot, you can take any number of AI models, from companies like OpenAI, Meta, Mistral and others and fine-tune them or point them to Fox News transcripts or Elon Musk’s tweets.

The AI era is about customization. Models are like broth, rather than soup. You can take a broth and make your own recipe, or you can buy a pre-made soup in a can. But you will be able to have any kind of soup you want.

Chris Lehane is OpenAI’s vice president of global affairs.

Reed Albergotti: What is in store for the next president of the United States on the AI front?

Chris Lehane: It really is how you strike and get to the right balance between ensuring that AI is deployed in a safe and responsible way, while also making sure that the country is poised to leverage its leadership role. For me, this ultimately all comes down to a simple construct: Whether democratic AI is going to prevail over autocratic AI.

This is a little bit strained at the technological level, but here’s a historical analogy: In 1944, the world gets together at Bretton Woods. There’s a conversation about how we can create a global financial system. Ultimately we decided to build that on the US dollar, with the idea that the dollar would help promote and advance “small d” democratic values around the world. There is a way of thinking about AI that is similar at this moment in time.

AI summits have taken place in London, Seoul, and elsewhere. Could that be where this Bretton Woods scenario plays out?

Yeah, you were asking what the next president is going to have to deal with. There’s no question that a manifestation of that could be bringing different aspects of the world together to talk about how we develop that “small d” democratic AI.

If you take a step back and look out at the world today, there are different circles to think about. First, it’s the US. And then it’s the Five Eyes [intelligence alliance] countries. Then there’s NATO countries and then countries we work with in Asia. There’s a next circle after that which you can almost think about as “swing states,” in parts of Africa, the Middle East and other parts of the world. And from a commercial diplomacy perspective, those countries are going to have to choose between the US and China in terms of who they align with.

Ultimately, you would want to have as many of these countries as possible building on democratic AI and not autocratic AI. Those are some of the really big calls that the next administration is going to have to make.

I want to ask you about your recent New Yorker profile. I know everyone in Silicon Valley is obsessed with youth and plastic surgery, but they really seemed to be obsessed with your hair and teeth. What is that all about?

I sent it to my 18-year-old and when he read that thing — it took him a couple hours because 9,000 words is not exactly what the younger generation is reading these days — his only comment was in a text message. He called me “Baldy.”

Elon Musk made a surprise (videocall) appearance at Riyadh’s FII conference where he predicted that, over time, all countries will have their own AI clusters, similar to how they have electricity grids or national airlines, Semafor’s Sarah Dadouch reports.

Building an AI cluster remains very difficult, Musk told the room of business and political leaders. To train a frontier model, states will need “a massive amount of compute and a level of technical skill that only a few companies possess,” he said.

Musk also predicted a future in which there will be more humanoid robots than people — 10 billion by 2040 was his estimate. “Assuming we are on the good path of AI, I think we will be in a future of abundance,” he added.

“Basically, anyone will be able to have any goods and services that they want,” a very optimistic Musk said. “The actual marginal cost of goods and services will be extremely low in the future.”

The size of a new partnership between private equity firm KKR and Energy Capital Partners to speed the development of AI. The project will help build out the data centers, power and other infrastructure needed to support the sector, the two announced today. It follows a similar fund launched recently by Microsoft, BlackRock, and Global Infrastructure Partners, while SoftBank and Apollo have held talks on a separate partnership. The multi-billion dollar cash piles reflect how much capital will be needed to keep up with the development of AI and the ecosystem needed for those efforts.

“In order for the US to maintain its advantage in AI, we will need massive new investments in power infrastructure on an accelerated basis that are capable of addressing concerns related to electricity prices and carbon emissions,” said Doug Kimmelman, founder of Energy Capital Partners, which invests in renewable energy projects. They estimate data center demand will need about $1 trillion in investments by 2030.

“You don’t want all of your eggs in the basket of a Taiwan fab,” Intel CEO Pat Gelsinger said in May 2021. The simple, 14-word sentence landed Intel in hot water, Reuters reported in a lengthy article published Tuesday that raised questions about the storied company’s future.

Gelsinger was referring to the risk that China might invade Taiwan, but the statement offended TSMC, the semiconductor manufacturer that operates the world’s most important fab, or chip manufacturing facility. TSMC was giving Intel a 40% discount, but canceled the special deal after Gelsinger’s comments.

SoftBank CEO Masayoshi Son offered rare, albeit cryptic, details about a $150 million robot factory he is due to open in Saudi Arabia in December during a conversation at the Future Investment Initiative in Riyadh, Semafor’s Mohammed Sergie reported.

“It’s not just the robot, it’s the robot with AI, the robot with the brain. That’s what we do,” Son said.