Tech Brew Ride Home - Tue. 05/28 – GPT5 Still A Year Out?

Episode Date: May 28, 2024

We have a rough timeline of when we can expect GPT5 and it looks like it won’t be this summer. xAI has a big raise. Another in-depth look at Apple’s AI strategy. The surprising old school companies getting boosted by the AI boom. And more data on how popular ad-supported streaming is becoming. Sponsors: YahooFinance.com 1Password.com/ride Links: OpenAI Says It Has Begun Training a New Flagship A.I. Model (NYTimes) Elon Musk’s xAI raises $6 billion to fund its race against ChatGPT and all the rest (The Verge) Apple Bets That Its Giant User Base Will Help It Win in AI (Bloomberg) Google adds AI-powered features to Chromebook (TechCrunch) The future of financial analysis: How GPT-4 is disrupting the industry, according to new research (VentureBeat) AI Is Driving ‘the Next Industrial Revolution.’ Wall Street Is Cashing In. (WSJ) What Happened to Our Ad-Free TV? (NYTimes) Learn more about your ad choices. Visit megaphone.fm/adchoices

On April 4th, 2023, around 2 in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16. Welcome to the TechMeme right home for Tuesday, May 28th, 2024. I'm Brian McCalla today. We have a rough timeline of when we can expect GPT5, and it looks like it will not be this summer. XAI has a big raise, another in-depth look at Apple's AI strategy, the surprising old-school

companies getting boosted by the AI boom, and more data on how popular ad-supported streaming is becoming. Here's what you miss today in the world of tech. So it sounds like GPT-5 is maybe a year away still. The Times is reporting that Open AI has begun training its new flagship AI model, but that's a process that could take nine months or more. And they've created a new safety and security committee since the alignment team is now dead. Quote, the company said in a blog post that it expects the new model to bring, quote, the next level of capabilities as it strives to build artificial general intelligence or AGI, a machine that can do anything the human brain can do. The new model would be an engine for AI products, including chatbots, digital assistance akin to Apple's Siri, search engines, and image generators. OpenAIA also said it was creating a new safety and security committee

to explore how it should handle the risks posed by the new model and future technologies. While we are proud to build and release models that are industry leading on both capabilities and safety, we welcome a robust debate at this important moment, the company said. Digital training of AI models can take months or even years. Once the training is completed, AI companies typically spend several more months testing the technology and fine-tuning it for public use. That could mean that OpenAI's next model will not arrive for another nine months to a year or more. As OpenAI trains its new model, its new Safety and Security Committee will work to hone policies and processes for safeguarding the technology the company said. The committee includes CEO Sam Altman, as well as OpenAI board members Brett Taylor, Adam DeAngelo, and Nicole Seligman.

The company said that the new policies could be in place in the late summer or fall. Open AI has folded its long-term safety research into its larger efforts to ensure that its technologies are safe. That work will be led by John Schulman, another co-founder who previously headed the team that created ChatGPT. The new Safety Committee will oversee Dr. Schulman's research and provide guidance for how the company will address technological risks, end quote. XAI, Elon Musk's AI play, has raised a $6 billion Series B from Valor, ViCapital, A16Z, Sequoia, Saudi Arabia's kingdom holding and others. Elon Musk says this is at an $18 billion pre-money valuation. That sounds about right. As I've said before, this deal was being shopped around by a ton of people to a ton of people, including myself. I don't know why, but there was a lot of allocation giving to a lot of

SPVs out there. We did not end up participating in this investment, just FYI, quoting the verge. Elon Musk founded XAI last summer, and today it announced raising $6 billion in funding, saying it will help bring these startups first products to market, build advanced infrastructure, and accelerate the research and development of future technologies. So far, XAI has launched GROC, a supposedly edgier version of OpenAI's chat GPT, available via X, formerly known as Twitter, where the chatbot is currently only available to X premium subscribers. The hardware capable of powering AI development is pretty pricey with NVIDIA's upcoming

Blackwell B-200 AI graphics cards costing anywhere from $30,000 to $40,000 apiece. Last week, a report from the information said that XAI would need 100,000 of Nvidia's current H-100 chips for a supercomputer to power an upgraded version of its GROC AI chatbot. Elon Musk reportedly told investors the plan is to launch the new data center by the fall of 2025. Outside of XAI and OpenAI, Musk said he would, quote, prefer to build products outside of Tesla when it comes to AI and robotics, unless he gets more control. Tesla shareholders will start voting this week on whether to restore Musk's $56 billion pay package ahead of its annual meeting on June 13th, end quote.

Meanwhile, Mark German has more on Apple's AI strategy and what we can maybe expect from WWDC next month. Quote, at the heart of Apple's new strategy is Project Grey Matter, a set of AI tools that the company will integrate into core apps like Safari, photos, and notes. The push also includes operating system features such as enhanced notifications. The system will work as follows. Much of the processing for less computing-intensive AI features will be run entirely on the device. But if a feature requires more horsepower, the work will be pushed to the cloud.

Apple is bringing the new AI features to iOS 18 and MacOS 15, and both operating systems will include software that determines whether a task should be handled on the device or via the cloud. Most of the on-device features will be supported by iPhone, iPad, and Mac, chips released in the last year or so. The cloud component, meanwhile, will be powered by M2 Ultra Chips located in data centers, as I've previously reported. There are several new capabilities in the works for this year, including ones that transcribe voice memos, retouch photos with AI, and make searches faster and more reliable in the spotlight feature. They also will improve

Safari web search and automatically suggest replies to emails and text messages. The Siri personal assistant will get an upgrade as well with more natural sounding interactions based on Apple's own large language models, a core technology behind generative AI. There's also a more advanced Siri coming to the Apple Watch for on-the-go tasks. Developer tools including Xcode are getting AI enhancements too. One standout feature will bring generative AI to emojis. The company is developing software that can create custom emojis on the fly based on what users are texting. That means you'll suddenly have an all-new emoji for any occasion beyond the catalog of options that currently offers on the iPhone and other devices. Another fun improvement, unrelated to AI, will be

the revamped iPhone home screen that will let users change the color of app icons and put them wherever they want. For instance, you can make all your social icons blue or finance-related ones green, and they won't need to be placed in the standard grid that has existed since day one in 2007. Another big part of the AI effort is creating smart recaps. The technology will be able to provide users with summaries of their missed notifications and individual text messages, as well as of web pages, news articles, documents, notes, and other forms of media. Apple is considering marketing the capabilities as a preview, at least in developer beta versions before a formal launch in September, indicating that the technology isn't yet fully baked, end quote.

Google this morning brought new AI features to Chromebook Plus devices, saying new Chromebook plus users will get a free year of Google One AI premium. Quoting TechCrunch. As a reminder, Chromebook Plus laptops are productivity-focused devices with minimum spec requirements. The company also unveiled new Chromebook Plus models from HP, Acer, and Asus. The search giant also added plenty of Gemini-powered features to maps and calls for spam detection during its Google I.O. Developer conference earlier this month. Google is now placing the Gemini icon on the app shelf of Chromebooks for easy access. What's more, users buying a new Chromebook Plus will get the Google One. AI premium plan for 12 months. This plan allows them to access Gemini Advanced, 2 terabytes of storage, and Gemini in docs, sheets, slides, and Gmail. Earlier this year, Google introduced a

writing tool within Chrome to help people write or reformat text anywhere. The company is now adding this ability to rewrite text system-wide on Chromebook Plus devices. Users can select text on PDFs, websites, or web apps to help them reshape the tone and length. Google is also adding a tool to Chromebook Plus for folks to generate wallpapers and video conferencing backgrounds through guided prompts. Last year, the company also added a generative AI-powered wallpaper tool to Android 14. Chromebook Plus devices are also getting a generative AI-powered magic editor in Google Photos. This feature lets you shift or resize, selected objects, and AI takes care of background regeneration based on the changes, end quote.

Researchers have found that GPT4 can outperform human analysts in predicting the direction of future corporate earnings, even when the AI was only given financial statements, quoting Venture Beat. The findings published in a working paper titled Financial Statement Analysis with Large Language Models could have major implications for the future of financial analysis and decision-making. We find that the prediction accuracy of the LLM is on par with the performance of a narrowly trained state-of-the-art ML model. The author's right, LLM prediction does not stem from its training memory. Instead, we find that the LLM generates useful narrative insights about a company's future performance, end quote. A key innovation was the use of chain of thought prompts that guided GPT4 to emulate the analytical process of a financial analyst,

identifying trends, computing ratios, and synthesizing the information to form a prediction. This enhanced version of GPD4 achieved a 60% accuracy in predicting the direction of future earnings, notably higher than the 53-57% range of human analysts forecasts. Taken together, our results suggest that LLMs may take a central role in decision-making, the researchers conclude. They note that the LLM's advantage likely stems from its vast knowledge base and ability to recognize patterns and business concepts, allowing it to perform intuitive reasoning even with incomplete information. The findings are all the more remarkable, given that numerical analysis has traditionally been a challenge for language models.

One of the most challenging domains for a language model is the numerical domain, where the model needs to carry out computations, perform human-like interpretations and make complex judgments, said Alex Kim, one of the study's co-authors. While LLMs are effective at textual tasks, their understanding of numbers typically comes from the narrative context, and they lack deep numerical reasoning or the flexibility of a human mind. Some experts caution that the AN model used as a benchmark in the study may not represent the state-of-the-art in quantitative finance. That ANN benchmark is nowhere near state-of-the-art, commented one practitioner on the Hacker News Forum. People didn't stop

working on this in 1989, they realize they can make lots of money doing it and do it privately, end quote. As AI continues, its rapid advance, the role of the financial analysts may be the next to be transformed. While human expertise and judgment are unlikely to be fully replaced anytime soon, powerful tools like GPT4 could greatly augment and streamline the work of analysts, potentially reshaping the field of financial statement analysis in the years to come, end quote. AI demand is driving a rally in old school stocks in the utilities, energy, material sectors, which are needed to make and operate AI products. Quoting the journal, the utility sector of the S&P 500 has returned 15% over the past three months, topping all

other corners of the index. Energy and material stocks have outperformed the broader market, which has advanced 4.2% over that period. Share prices are surging for industrial firms that stand to benefit from data center expansion and renovation. That is where the investable opportunities are, said Lauren Goodwin, economist, and chief market-strand. at New York Life Investments. It's data center builders and operators, power and utilities, she said. Data Center executives have said that surging demand means longer delivery times for essential equipment such as backup generators and cooling systems. We have seen a broadening out of the AI trade. It's no longer dependent on just one stock, said Nadia Lovell, senior U.S.

Equities Strategist, Global Wealth Management at UBS. Chips are foundational, but they're not the whole house, end quote. The global ex-U.S. infrastructure. Development ETF has returned 13% this year, outpacing the S&P 500's 11% advance. Shares of Verteev Holdings, which makes equipment that powers and cools data centers have more than doubled this year. The company said new orders grew 60% from a year earlier in the first quarter. Wall Street analysts had expected about 15% growth and have revised their profit forecasts sharply higher for the years to come. Though still in its early stages, AI is quickly becoming a pervasive theme across R&Markets, Veritiv's CEO,

Diorando Albertazzi said last month, Eaton, a company that makes power management equipment, has advanced 42% this year. Johnson controls which manufactures electronic systems for commercial buildings is up 28%. End quote. Finally today, some non-AI stuff. The Times has a look at the growing importance of ads for video streaming services,

as Antenna says that 56% of new subscribers chose the cheaper ad-supported tier in Q1. which lines up, I believe, with what Netflix told us recently. Quote, consumers can still avoid most of the ads for a price. Most streaming services still have an ad-free version, including Amazon, which requires subscribers to pay an extra $3 a month to skip the ads. Apple TV Plus continues to offer only an ad-free experience. The commercial tiers, however, are becoming more essential to the overall streaming business.

There were at least 93 million ad-supported streaming subscriptions in the United States at the end of last year, according to estimates from Brian Weiser and industry analyst and Antenna, a subscription research firm. In the wake of Amazon's automatic switch to advertising and more ad-tier customers picked up by other streaming services, Mr. Weiser and Antenna estimate that there are at least 170 million ad-supported subscriptions now. Through the first three months of 2024, 56% of new subscribers to a streaming service chose the lower-price ad-tier, according to Antenna. That was up from 39% a year earlier, the firm said. executives have tried to assure subscribers that while advertising is back, it won't be as overwhelming as in traditional television.

Just a few years ago, an episode of a prestige, basic cable drama like Ryan Murphy's American Crime Story, was interrupted by 21 minutes of commercials, but ads take up far less time on streaming services. For instance, on Disney Plus, the average amount of time for commercials is four minutes per hour. On Hulu, it's just over six minutes. There was always this notion that people don't like ads, said Rita Farrow, the president of ad sales at Disney, I don't think that's true. People don't like bad advertising or a bad advertising experience, she said. In the data-rich streaming world, she argued the advertising experience is better informed than it was on traditional television, and the company knows what a person's viewing preferences are,

end quote, what products are relevant to you, she said, end quote. Nothing more for you today. Talk to you tomorrow.