Silicon Valley’s Swing Towards Trump

[00:00:00] Clawed by Anthropik is generative AI for everyone. Clawed 3.5 Sonnet offers groundbreaking intelligence at a price point that works for heavy-duty tasks. Join thousands of enterprises using Anthropik to stay at the frontier of AI. Learn more at anthropik.com slash clawed.

[00:00:22] Welcome to Tech News Briefing, it's Thursday, July 18th. I'm Zoe Thomas for The Wall Street Journal. Big names in tech are putting their support, in their money, behind Donald Trump's presidential campaign. We'll tell you whether it signals a brewing culture change in traditionally liberal Silicon Valley.

[00:00:42] And then, the Chinese government is putting a lot of resources behind the country's generative artificial intelligence companies as they work to gain ground on US competitors. But China's political controls threaten to weigh down those efforts. We'll explain how that could affect China's position in the AI race.

[00:01:06] But first, earlier this week we reported that Elon Musk has told people he will donate around $45 million a month to a pro-Trump super political action committee called America PAC. Other big Silicon Valley names say they plan to donate to organizations supporting Trump too.

[00:01:25] It seems to signal a shift in the historically liberal home of the tech sector. Here to tell us more about this is our reporter, Dana Madioli. Dana, what do we know about why Musk is putting so much money into America PAC?

[00:01:39] Elon Musk, we've noticed this political shift with him over the last year or so. In prior elections, he's been an independent or he's acknowledged that he's voted for Democrats and we've seen him start to express angst with the Democratic Party under President Biden.

[00:01:57] He hasn't agreed with some of the more liberal leaning ideals like diversity, equity and inclusion. He has felt like the Biden administration has slighted him when it comes to Elon's role at Tesla and how they've revolutionized this electric vehicle movement and we've seen him adopt more right-wing views.

[00:02:17] So Dana, he's not the only big name from Silicon Valley to be putting money towards Trump's campaign. Who else is? We've noticed this broader shift in the valley, which has historically been very liberal leaning.

[00:02:30] When we got the documentation for the America PAC, there were some other high-profile San Francisco names on the list. Some of them include Joe Lonsdale. He is the co-founder of Palantir. He's also a friend of Elon Musk's.

[00:02:44] Antonio Gracius, who is a former Tesla director and he's currently on the board of Musk's SpaceX company. Ken Howary, which was a member of the original PayPal mafia as it's called, which is the diaspora of all these PayPal executives where Musk was as well.

[00:03:01] And then also two very high-profile venture capital investors from Sequoia, Doug Leone and Sean McGuire. Mark Andreessen and Ben Horowitz, the founders of VC firm Andreessen Horowitz told staff on Monday that they intend to contribute personal money to support pro-Trump political organizations.

[00:03:22] And that's according to people with knowledge of the meeting. What did they say about that decision? So on that front Andreessen had endorsed Hillary Clinton in the 2016 presidential election. And at the time he said that he opposed Trump's sances on immigration. In 2020, he didn't endorse a candidate publicly.

[00:03:40] So, you know, we've seen a shift in him as well. And in a blog post earlier this month, he said that his political efforts were focused on protecting startups, which he described as little tech and he wanted to protect them from government regulation.

[00:03:54] Broadly, what issues in particular have folks in tech supporting Trump over Biden? It really ranges the gamut. We've spoken to a lot of people about this. For some people they think that the Democratic Party has become too woke

[00:04:07] and they oppose some of these DEI efforts and the like. Other people do not like Biden's handling of Israel and Gaza. Other people we've spoken to have criticized Biden's stance on artificial intelligence regulation and all these crackdowns on acquisitions by Big Tech.

[00:04:25] Trump's pick for Vice President Ohio Senator J.D. Vance has ties to Silicon Valley. What are they and how might that help the Trump campaign with the tech sector? After J.D. Vance wrote his book, Hillbilly Elegy, Peter Thiel hired him at his venture capital firm, Founder's Fund.

[00:04:43] So he's a bit of a known entity in Silicon Valley. He worked as a venture capitalist. He wound up later opening his own VC firm, which invested in companies in Middle America, tech companies.

[00:04:55] When Trump was hosted for a big fundraiser in San Francisco last month by David Sacks, another investor, J.D., was alongside with him and introduced him to the crowd and he's friends with a lot of the people that were in that room. That was our reporter Dana Madioli.

[00:05:09] Coming up, China is putting the power of the state behind AI development and it risks strangling it. That story after the break. China is turning to an old playbook to compete in the artificial intelligence race, putting the vast resources of the state behind Chinese companies.

[00:05:58] At the same time, the heavy hand of the Beijing government is threatening to hobble its AI ambitions. Here to tell us more about the situation this is creating is our China tech reporter and editor, Lisa Lin. So Lisa, lay the groundwork for us.

[00:06:12] Where do Chinese companies and I guess China more broadly stand in the AI race? It's an interesting one. China started out ahead of the pack when it came to AI because if you think about AI, it's not just generative AI.

[00:06:26] It's everything from computer vision to natural language processing and when it came to computer vision, which was really kind of what sparked the world's interest in AI in the first place, that was where China was ahead of the pack.

[00:06:38] China was training models that ended up in surveillance cameras. So a lot of surveillance cameras you see around the world now have AI models that can spot items or people's faces and a lot of that started in China. However, when it's come to generative AI,

[00:06:54] we've seen China take a bit of a backseat in this race. We've seen US models really pull ahead of the pack and open AI's chat GPT is currently universally acknowledged that it's in the lead. What kind of resources is the government putting behind companies developing AI?

[00:07:11] Basically to catch up with the West now, China's using a very familiar approach that they've used in several other industries. It's an industrial policy approach that's very top-down in which they put a lot of the state's resources to developing industry.

[00:07:26] They've done it with EVs, they've done it with batteries and now they're doing it with AI. So what they're doing with AI is trying to help ease up the bottlenecks on two fronts. One is on the compute power. China's been blocked by the US from acquiring high-end chips

[00:07:40] that help AI companies run and train their algorithms and to get over that hump, they're pooling together the scarce resources of high-end chips they actually have in China now. They're building large data centers and they're offering this computing power to companies at a subsidized cost.

[00:07:57] So that's how they're dealing with the compute. With data because there isn't that much Chinese language data out there, not on the web and in China, a lot of it's, frankly, behind some firewall or protected by a company. So the Chinese government is helping to build data sets

[00:08:13] that Chinese AI startups could use to train their large language models on. Let's take those two elements one at a time. You mentioned that China is working to develop its own AI chips. Can you tell us more about that?

[00:08:26] When it comes to domestic chips, it seems like the front runner in this race is Huawei. Huawei's come up with several chips which have ended up in a lot of these data centers that I talked about earlier.

[00:08:37] And China has been backing Huawei as well with a lot of subsidies. When it comes to the data, you're reporting there's also tight restrictions that are affecting this. Can you tell us how Chinese companies are training their large language models given those rules?

[00:08:52] Yeah, so back up a bit to explain what are the rules governing data in China. The biggest issue in China right now is the censorship. So online on the Chinese Internet, a lot of the Internet discourse has to be censored and filtered so that it doesn't run afoul

[00:09:07] of Communist Party kind of guidelines. And that can be anything from why did the current Chinese president Xi Jinping seek a third term which is out of the norm or even just content on Tiananmen Square which was an incident in the 80s in which the Chinese army

[00:09:24] fired on Chinese students and protesters. So there are a lot of these content rules that restrict what you can say in China and the same rules apply for large language models and chatbots like AI, for example. So your chatbot has to be regulated by China

[00:09:42] and actually this is a real problem for AI startups in China because not only do you have to overcome like the bottlenecks of compute and data, you still have to make sure that the chatbot or the large language model that you develop is consistent with these content rules.

[00:09:58] And how does that process work to make sure it's consistent? What the Chinese Internet regulator says is your large language models have to be approved before they're released to the public. In practice what we've found is there's this whole cottage industry

[00:10:13] of consultants that have popped up to help ensure that your large language model has a smooth passage through that approval and what happens is before you actually get approval you have to run by that chatbot at least 20,000 questions on various topics that could touch on red lines

[00:10:34] that the Communist Party feels might be offensive and the chatbot has to answer in a way that it doesn't cross those red lines. Did the cyberspace administration of China which gives the approval for these AI models comment for your story? Without reach them and they did not comment.

[00:10:53] What are the people in the industry that you spoke to expect the outcome of China's approach to be? The most consistent thing is a lot of people would think that the lack of compute power will ultimately hamstrung China's AI industry in the end.

[00:11:07] The lack of access to tie-in chips will ultimately get to Chinese AI companies because your large language models will ultimately require more and more compute to train and to be finessed. So most industry analysts expect the gap between China and the US to ultimately start to grow.

[00:11:26] That was China Tech Reporter and Editor Lisa Lin. And that's it for Tech News Briefing. Today's show was produced by Julie Chang with Supervising Producer Catherine Milsoff. We had additional support from today's Ruiz Sandoval. I'm Zoe Thomas for The Wall Street Journal.

[00:11:42] We'll be back this afternoon with TNB Tech Minute. Thanks for listening. Clod by Anthropik is AI for everyone. The latest model, Claude 3.5 Sonnet, offers groundbreaking intelligence at a price that works for heavy-duty tasks. Claude Sonnet can generate code, help with writing,

[00:12:07] and reason through hard problems better than any model before, pushing the limits of what your business can accomplish with AI. Join the thousands of businesses that choose Anthropik to stay at the frontier of AI. Remember how Claude can transform your business at anthropik.com slash Claude.