More Than 1/4 of Google's Code Now Generated by AI

0:00Today on the AI Daily Brief, Google reveals that more than a quarter of their new code is now generated by artificial intelligence. Before that, in the headlines, Elon's XAI is apparently raising new funding at a$40 billion valuation. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. To join the conversation, follow the Discord link in our show notes. Welcome back to the AI Daily Brief Headlines Edition, all the daily AI news you need in around five minutes. We kick off today with the latest funding rumors. In the wake of OpenAI's massive$157 billion valuation round, Elon Musk's XAI is looking to raise billions as well, at a valuation that if not the stratospheric$157 billion is still very high.

0:46According to Wall Street Journal reporting, the company is in talks to raise at a$40 billion valuation. The news comes just five months after XAI's previous round, which raised$6 billion on a$24 billion valuation. Since then, the company has completed construction of the Colossus supercomputer at record speed and rolled out multiple new features, meaning that to some investors, the jump in valuation is likely justified. The journal emphasized that the deal is still in its early stages, so is subject to change or abandonment. Musk's plans for the money are pretty clear. On a video called Tuesday into a conference in Saudi Arabia, Musk said, If you're training a frontier model, you need a massive amount of compute.

1:19On Monday, he confirmed plans to double the size of the Colossus training cluster by adding an additional 100 ,000 GPUs. Nikolai Yakovenko writes, Elon wants a true OpenAI competitor. And say what you will, but Elon's companies rarely fail to raise the money they ask for. The Elon Facts account, which I will let you decide what its likely bias is, pointed out that NVIDIA CEO Jensen Huang recently said, Building a massive supercomputer factory in the short time that was done, that is superhuman. There's only one person in the world who could do that. What Elon and the XAI team did is singular. Never been done before.

1:53Everard AI's Jai Azang writes, Underrated, XAI and Google. Accurately rated, OpenAI, Anthropic and Meta. And I actually think that the bigger point here is not so much the underrated or accurately related designation, but the fact that when it comes to frontier models, if you are actually trying to compete in that space, if you are an investor who has a thesis that the value proposition of AGI is going to be so large that effectively any price is worth it to invest in, these plus Mistral are pretty much the only way you can make that bet, which means that the price is fairly inelastic. Now, we had previously gotten reports that Anthropic had also floated$40 billion in its own funding talks, and my guess is that if Elon is coming anywhere near close to that, Anthropic's going to use that fact to try to go for something even higher than$40 billion, making the argument that they are proportionally farther ahead than Elon is.

2:43Who knows, though? We will have to wait and see. Meanwhile, some interesting hardware news. OpenAI will reportedly have their own custom-designed AI chip ready to use by 2026. According to Reuters, the company has explored all options to diversify their chip supply and reduce costs, considering building everything in-house and raising capital to build a network of chip foundries, but now it appears that they've abandoned those plans due to cost and time constraints. Instead, according to Reuters sources, they are opting to focus on in-house chip designs. The plan will now involve partnering with Broadcom or TSMC for manufacturing while adding AMD alongside NVIDIA as a supplier.

3:16The first AMD installation will come through Microsoft Azure, who are using the company's MI300X chips. Those chips are not as performant as NVIDIA's H100s in training, but have outperformed in inference benchmarks. Sources said that OpenAI had already been working with Broadcom for months to produce their first inference chip. Currently, GPU supply has been bottlenecked by strong demand for training chips, but there's a chance that inference will become the scarce resource as more AI applications are deployed. OpenAI have reportedly assembled a 20-person chip design team led by top former Google engineers Thomas Norrie and Richard Ho.

3:47Sources also say that the firm has secured manufacturing capacity from TSMC in 2026. OpenAI CFO Sarah Fryer referenced the growing pains as the company learns to create its own data centers and chips during an interview earlier in the week, stating, It's definitely a stretch from a capital perspective, but also my own learning. Frankly, we are all learning in this space, infrastructure is destiny. Salesforce, which is making a huge bet on agents, has finally released them broadly via their new agent development platform, AgentForce, to the public. The product is essentially a low or no-code way of building and deploying agentic chatbots for employees and customers.

4:21Salesforce outlined a key use case in a post highlighting the upgraded functionality, writing, Your customer at midnight asks, I need to change my entire order. Traditional bots say, I've created a ticket. Please wait 24 to 48 hours. AgentForce service agent, order updated, shipping rerouted, confirmation sent, anything else. In a press release, the company said, AgentForce doesn't depend on human engagement to get work done. These agents can be triggered by changes in data, business rules, or pre-built automations. Now, Salesforce CEO Mark Benioff has recently become the leading voice saying that the assistant era was all BS and that everything is about agents, which is clearly marketing given where they sit and a way to poke at Microsoft.

4:59But still, I think it's pretty exciting to start to see these agents actually rolling out in a real way. We'll soon be able to see a lot more around how well they work in practice. That, however, is going to do it for today's AI Daily Brief Headlines Edition, next up, the main episode. Today's episode is brought to you by Venice. Venice is a private, uncensored, generative AI app. It accesses open-source models to enable text, image, and code generation without the fear of being spied on or having your data exploited. Discuss anything with Venice without concern about it being monitored, sold, or given to advertisers and governments.

5:31Venice is different because your conversations and creations are kept securely within the browser, never stored or accessible by Venice. Unlike other AI apps, Venice won't tell you what's okay to say or not. Venice won't patronize you. It simply provides direct access to machine intelligence. No topics are off limits. No ideas are taboo. With Venice, you're in control of the AI as you should be. Pro subscriptions are available for $49 a year or$8 per month. AI Daily Brief listeners receive a 20 % discount on Venice Pro. Visit venice.ai slash nlw and enter the discount code nlwdailybrief. That's nlwdailybrief, all one word.

6:07Today's episode is brought to you by Superintelligent. Every single business workflow and function is being remade and reimagined with artificial intelligence. There is a huge challenge, however, of going from the potential of AI to actually capturing that value. And that gap is what Superintelligent is dedicated to filling. Superintelligent accelerates AI adoption and engagement to help teams actually use AI to increase productivity and drive business value. An interactive AI use case registry gives your company full visibility into how people are using artificial intelligence right now. Pair that with capabilities building content in the form of tutorials, learning paths, and a use case library.

6:46And Superintelligent helps people inside your company show how they're getting value out of AI while providing resources for people to put that inspiration into action. The next three teams that sign up with 100 or more seats are going to get free embedded consulting. That's a process by which our super intelligent team sits with your organization, figures out the specific use cases that matter most to you, and helps actually ensure support for adoption of those use cases to drive real value. Go to bsuper.ai to learn more about this AI enablement network. And now, back to the show. Welcome back to the AI Daily Brief.

7:20Today we are talking about what is an absolutely insane statistic. Yesterday, Google reported quarterly earnings, and on the analyst call, Google CEO Sundar Pichai was asked effectively about the effectiveness of AI, how much it was real, how much it was hype. And one part of his response was that he said that more than a quarter of all new code at Google is being generated by AI than reviewed and accepted by engineers. Now, there was a lot of other stuff on this call as well. AI, for example, is helping drive revenue at Google. Google's cloud business, which includes its AI infrastructure products, was up 35 % year-over-year to$11.4 billion.

7:57But this stat around how much Google is dogfooding AI was really the headliner. Pichai went on to say that AI for coding was, quote, boosting productivity and efficiency, and that this helps our engineers do more and move faster. I'm energized by our progress and the opportunities ahead, and we continue to be laser-focused on building great products. Of course, as Business Insider points out, as impressive as this number is, there may be some who get a little bit nervous hearing it. The new data from Pichai will surely have some employees wondering whether they're coding themselves out of a job.

8:26Company leaders have previously promised that AI isn't taking Googler's jobs yet, but the over 25 % figure is striking and underscores the benefits of improving this technology. A lot of the discussion on X slash Twitter was around the implication for developers. Adit Sheth writes, if you're a programmer hesitant about using AI, and trust me, some are, it's starting to look like resisting it might mean getting left behind. Rohan Paul writes, connecting the dots, Google laid off its entire Python team in April 2024 as part of a cost-cutting effort. Today, it's revealed on the earnings call that more than 25 % of all new code at Google is generated by AI.

9:00Software engineering will be disrupted in more ways by AI than we can see right now. He also went on to make a prediction, I think that the next 10 years is the last period that any human will ever write any code. Others tried to connect it to their own industries. Lawyer Daniel Lina Jr. said, If more than 25 % of new code at Google is generated by AI and then reviewed by engineers, I find it quite plausible that the same is possible in law if we do the work to build capable systems. There are differences in the ecosystems and we need to close those gaps. Dharaj Nambiar writes, First AWS and now Google are automating large amounts of their code generation with the use of Gen.AI.

9:33Software engineers that don't leverage AI extensively have one, maybe two years of shelf life. Investor Steven Sanofsky made the point that ultimately, users don't really care around how products were built, they just care that they work. He wrote, In the early days of the micro, computer software was written in assembly. Then C got introduced to the micro and all the industry was asking of each vendor was, will you move to C and when. That was because C was viewed as faster and easier to use and created less bugs. It was a modern, higher-level language. Soon the biggest vendors all announced that their next versions would be in C.

10:02This happened again in 1990 or so with the rise of Object Oriented. Soon vendors were claiming that their products were OO and thus imbued with all the magic beans that came from that. They would be easier to maintain, have fewer bugs, easier to add new features. Nope, that wasn't true either. Then everyone began to doubt those techniques. No one cares what tools companies use to write code, not consumers or even enterprise. They care about features, cost, quality, security, reliability, privacy, and performance, to name a few. If new tools help, then great. If new tools don't help, then that's not good.

10:28It's even worse if the companies are out there touting new tools and not delivering. Still, Ravu Tanuku summed up the point of Sindar's comment, saying, This says to Wall Street we are getting leverage on AI spending. It portends what less sophisticated tech companies can accomplish over time. It implies ROI is happening and gives a reason why they should keep spending on it. I will take a minute here to get up on my bully pulpit once again and to talk about the two phases of generative AI adoption. Once again, there is inevitably going to be a phase where companies treat AI as exclusively an efficiency technology, a way to get the same output with less input, a cost-saving technology that will in many cases lead to job cuts, lower overall costs, and the markets are likely to reward it at least in the short term.

11:09However, the companies that win in the generative AI era, I'm quite convinced, will be those who view generative AI as an opportunity creation technology, A way, in other words, to do more with the same, or much, much more with just a little more. When it comes to developers, I am firmly on the other side of, in 10 years we won't have any humans writing any code. Or at least I'm on the other side when it comes to humans producing code. I think that we are going to have hundreds and hundreds or thousands of times the code that we have now. I think more people are going to be producing code with the help of both assistants and agents, and I think that means more things are going to be built.

11:43But in the in-between, it's going to get weird. And even as someone who watches this every day, hearing that Google is already generating a quarter of its code with AI is fairly surprising. Two other interesting Google nuggets from the earnings call, one small and one bigger. The small one is that Google says that they won't ship agentic features until next year at the earliest. The feature, known as Project Astra, was previewed at the IO Developer Conference in May. It encompasses a range of functionality, including a smartphone app that can recognize the world through the camera, to AI assistants that can complete tasks autonomously.

12:13On that same earnings call, CEO Sundar Pichai said, Google is building out experiences where AI can see and reason about the world around you. Project Astra is a glimpse of that future. We're working to ship experiences like this as early as 2025. You might remember that the information had previously reported that Google was planning to ship their first agents known as Jarvis as early as December, but now that timeline seems a little bit unclear. Maybe the bigger deal is that Microsoft's GitHub co-pilot will now support models from Anthropic and Google alongside OpenAI. Users will be given the choice of model between Cloud 3.5 Sonnet, Gemini 1.5 Pro, as well as GPT-40, O1 Preview, and O1 Mini.

12:50GitHub CEO Thomas Domke said, There is no one model to rule every scenario, and developers expect the agency to build with the models that work best for them. It is clear the next phase of AI code generation will not only be defined by multimodal functionality, but by multimodal choice. Microsoft, of course, introduced GitHub Copilot in 2021, making it one of the first products to demonstrate the power of AI assistance. It had previously relied solely on models provided by OpenAI. The arrival of OpenAI's O1 models led GitHub to explore the idea of a drop-down menu to provide easy access to model options.

13:20Domke said at that point it felt like the right time to expand to other companies as well. He added, We're planning on extending that list in the future but have no partnerships to announce at this point. GitHub has also introduced a new automated code review feature, and next up on the feature list is a powerful app designing tool called Spark. The feature will allow developers to create app prototypes based on text prompts and then refine the designs from there. Domke said, For too long, there has been an unscalable barrier of entry separating the vast majority of the world's population from building software.

13:47With Spark, we will enable over 1 billion personal computer and mobile phone users to build and share their own micro apps directly on GitHub. Now, of course, the subtext and context of this announcement is both, one, what appears to be a potentially fraying relationship between Microsoft and OpenAI, and two, of course, the battle between OpenAI and Anthropics models when it comes to code assistance. Developer Nick Dobos writes, GitHub co-pilot adopting Quad and Gemini is the final nail in the coffin for base model companies. AI infrastructure is a difficult business to compete in, even if they are producing amazing technology.

14:19Application layer wins. Wrappers all day, every day. Super fascinating. I don't really know how much to read into this when it comes to that OpenAI relationship piece. It certainly does put evidence in the column of those who think that base level models are going to be commoditized. But no matter what, it's a pretty big shift and one that'll be worth watching over time. For now though, that is going to do it for today's AI Daily Brief. Until next time, peace.