Tech Brew Ride Home - Mon. 09/25 – Amazon Makes A Big Bet On Anthropic

Episode Date: September 25, 2023

Amazon makes a big bet on Anthropic. A big DeFi hack. The US government is weighing a sort of “know your customer” rule for big cloud providers. OpenAI has rolled out some cool new ways to interact with their AI. And why the number of smartphone brands around the world has basically collapsed. Sponsored: NPR Planet Money Zbiotics.com/ride and code: ride Links: Amazon to invest up to $4bn in AI start-up Anthropic (Financial Times) Defi Project Mixin Network Suspends Services After $200 Million Crypto Hack (Bloomberg) Booking to appeal after EU vetoes $1.7 bln ETraveli deal (Reuters) White House could force cloud companies to disclose AI customers (Semafor) You can now prompt ChatGPT with pictures and voice commands (The Verge) WGA and the studios reach tentative deal to end writers’ strike (Los Angeles Times) Nearly 500 Brands Exited Smartphone Market During 2017-2023 (Counterpoint) 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 Tech Mean Bread Home for Monday, September 25th, 2023. I'm Brian McCullough today. Amazon makes a big bet on Anthropic, a big defy hack. The U.S. government is weighing a sort of know-your-customer rule for big cloud providers. OpenAI has rolled out some cool new ways to interact with their AI and why the number of smartphone brands around the world has basically collapsed. Here's what you miss today in the world of tech. Amazon is investing up to $4 billion in Anthropic. with an initial investment of $1.25 billion for a minority stake in the AI startup, but also with an option to increase their total investment to $4 billion,

quoting the Financial Times. It is a bid to forge a close relationship with a prominent AI startup akin to Microsoft's alliance with OpenAI, the group behind ChatGPT. As part of the agreement, Anthropic will use Amazon's cloud computing platform and its dedicated AI chips to create its models. AI startups have become locked in an arms race to secure, the costly chips and the data center resources necessary to build the latest AI systems called large language models. Anthropic was last valued at nearly $4 billion in a funding round

earlier this year, according to a person familiar with the terms. It is one of the top competitors to Open AI raising huge sums this year, including Inflection AI, which raised $1.3 billion from Microsoft and NVIDIA, and Toronto-based Cohere, which raised $270 million from Nvidia and others. The startup's new alliance with Amazon Web Services appeared to be a shift away from Google, which had invested $300 million in Anthropic last year. It comes just seven months after Anthropics said it would train its models on Google's chips and use its cloud. For Amazon, the deal marks its latest push to capitalize on the excitement around Generative AI, a technology capable of creating human-like text and realistic images.

Amazon is looking to position its Traneum and Inferentia chips as credible alternatives to NVIDIA's processors for the training and running of Generative. AI models. Anthropics clawed chatbot, which rivals ChatGBTGPT, is already among a range of AI products available on AWS's Bedrock Service, which allows customers to build generative AI applications in the cloud. The New Deal represents a, quote, significant expansion of the partnership with Anthropics said Adam Slipsky, head of AWS. Anthropic would have access to, quote, significant quantities of traneum chips to train future versions of its foundation models, he added, end quote. It's been a while since we've had one of these, or at least one big enough, I figured it was worth mentioning.

Defy Project Mixon Network has suspended deposits and withdrawals after suffering a hack involving about $200 million in crypto assets. Quoting Bloomberg, the breach was caused by a compromise in the project's cloud service providers database, according to blockchain security firm SlowMist, which is assisting Mixon in the investigation. Regarding how to deal with the lost assets, the Mixon team will announce the solution afterward, Nixon wrote in a blog post. Mixon Network aims to help blockchains handle more transactions by processing more with zero fees, according to its white paper. The top 100 assets on the network were valued at over $1.1 billion, according to its July monthly report. The project's native token, X-I-N, was trading at $195, down 8.6% over the past 24 hours, according to data from Coin Gecko,

end quote. The EU is officially blocking booking.com's 1.6 billion euro acquisition of E. Travelli, saying booking failed to allay concerns that the deal would bolster its dominance in the hotel online travel agency market, quoting Reuters. The European Commission, which acts as the competition watchdog for the 27-country European Union, said bookings remedies were not sufficient to address its concerns, confirming a Reuters story earlier this month. The veto suggests that EU regulators may be taking a tougher stance on mergers in the digital sector. EU antitrust chief Didier Rangers said the travel market was becoming more digital, in particular the online travel agency or OTA market, which includes hotels, flights,

car rentals, and attractions. Hotel OTAs are worth about 40 billion euros annually, the largest and most profitable segment of the OTA sector. The combination of network effects and consumer inertia means that many consumers get information on room availability and prices only from OTAs, often only from booking. Reinders told a press conference. He said the commission had reached out to almost 15,000 hotels for feedback on the proposed deal. Overall, market participants were concerned that the transaction would strengthen booking's dominant position on the market for hotel OTAs in the European economic area, reduce competition and increased prices for hotels and possibly for consumers, end quote. Booking said it would challenge the veto, which has one unconditional approval in the

United States and Britain. It also extended a flight agreement with E. Trevelli to 2028, underlining its determination to grow its flight business despite the EU ruling, end quote. This is taking things in a new direction. Sources say the White House is weighing requiring cloud companies to disclose when a client buys computing resources above a certain threshold as part of an possibly upcoming executive order, quoting semaphore. The provision would direct the Commerce Department to write rules forcing cloud companies like Microsoft, Google, and Amazon to disclose when a customer purchases computing resources beyond a certain threshold. The order hasn't been finalized, and specifics of it could still change.

Similar, know-your-customer policies already exist in the banking sectors to prevent money laundering and other illegal activities, such as the law-mandating firms to report cash transactions exceeding $10,000. In this case, the rules are intended to create a system that would allow the U.S. government to identify potential AI threats ahead of time, particularly those coming from entities in foreign countries. If a company in the Middle East began building a powerful large language model using Amazon Web Services, for example, the reporting requirement would theoretically give American authorities an early warning about it. The policy proposal represents a potential step

toward treating computing power or the technical capacity AI systems need to perform tasks like a national resource. Mining Bitcoin, developing video games, and running AI models like chat GPT all require large amounts of compute. If the measure is finalized, it would be a win for organizations like OpenAI and the Rand Corporation Think Tank, which have been advocating for similar know-your-customer mechanisms in recent months. Others argue it could amount to a surveillance program if not implemented carefully. The details are really going to matter here, said Clon Kitchen, a non-resident senior fellow at the American Enterprise Institute, where he focuses on national security and emerging technology. I understand why the administration is trying to get at this

issue, we're going to need a strategic understanding of adversarial development of these models, end quote. One major challenge for this approach, the amount of computing power it takes to build powerful models like chat GPT is rapidly falling, thanks to improvements in the algorithms used to train them. By the time the Commerce Department decides on a reporting threshold, it could already be out of date, and trying to make effective updates will be like chasing a moving target, end quote. OpenAI is rolling out an update to let Plus and Enterprise users prompt ChatGPT using voice commands or by uploading an image available for other users soon after. Quoting the Verge, most of OpenAI's changes to ChatGPT involve what the AI powered bot can do,

questions it can answer, information it can access, and improved underlying models. This time, though, it's tweaking the way you use ChatGPT itself. The company is rolling out a new version of the service that allows you to prompt the AI bot not just by typing sentences into a text box, but by either speaking aloud or just uploading a picture. The new features are rolling out to those who pay for chat GPT in the next two weeks, and everyone else will get it soon after, according to OpenAI. The voice chat part is pretty familiar. You tap a button and speak your question. ChatGPT converts it to text and feeds it to the large language model,

gets an answer back, converts that back to speech, and speaks the answer out loud. It should feel just like talking to Alexa or Google Assistant, only OpenAI hopes. The answers will be better thanks to the improved underlying tech. It appears most virtual assistants are being rebuilt to rely on LLMs. OpenAI is just ahead of the game. OpenAI's excellent whisper model does a lot of the speech-to-text work, and the company is rolling out a new text-to-speech model it says can generate human-like audio from just text and a few seconds of sample speech.

You'll be able to choose ChatGPT's voice from five options, but OpenAI seems to think the model has vastly more potential than that. OpenAI is working with Spotify to translate podcasts into other languages, for instance, all while keeping the sound of the podcaster's voice. There are lots of interesting uses for synthetic voices, and OpenAI could be a big part of that industry. But the fact that you can build a capable synthetic voice with just a few seconds of audio also opens the door to all kinds of problematic use cases. These capabilities also present new risks, such as the potential for malicious actors to impersonate public figures or commit fraud. The company said in a blog post announcing the new features, the model isn't available. for broad use for precisely that reason. OpenAI says it's going to be much more controlled and

restrained to specific use cases and partnerships. The image search, meanwhile, is a bit like Google lens. You snap a photo of whatever you're interested in, and chat GPT will try to suss out what you're asking about and respond accordingly. You can also use the app's drawing tool to help make your query clear or speak or type questions to go along with the image. This is where chat GPT's back and forth nature is helpful. Rather than doing a search, getting the wrong answer, and then doing another search, you can prompt the bot and refine the answer as you go. This is a lot like what Google is doing with multimodal search too, end quote. So again, we knew this. What, a year ago, that all you had to do was, you know, upload a couple episodes of this show and you could train it

to be exactly my voice. But now the thing that Spotify is partnering with Open AI to do is what we were saying before about AI-powered stuff at scale, because this will now allow certain Spotify to translate their podcasts immediately into different languages, but maintain, you know, the humble podcaster's own voice. It's pretty wild. You might have heard that the WGA and AMPTPP have reached a tentative deal to end the whole writer's strike in Hollywood. Sources say the proposed three-year contract adds new AI rules, increases streaming residuals, and more, and since those are tech concerns that we talk about a lot streaming and AI stuff and they were key to this strike happening, I thought it was worth looking at where this all settled, quoting the Los Angeles Times. The proposed three-year contract, which would still have to be ratified by the union's 11,500 members, would boost pay rates and residual payments for streaming shows and impose new rules surrounding the use of artificial intelligence. The writer's strike was in many ways a response to the tectonic changes wrought by streaming. Shorter seasons for streaming shows and fewer. are writers being hired have cut into guild members' pay and job stability, making it harder to earn a

sustainable living in the expensive media hubs of Los Angeles and New York, guild members have said. The studios came into negotiations with their own set of challenges, though. The pay TV business is in decline because of cable court cutting and falling TV ratings, which have eroded vital sources of revenue. At the same time, the traditional companies have spent massively to launch robust streaming services to compete with Netflix, losing billions of dollars in the process, end quote. And then quoting the New York Times. The use of AI was the final sticking point. The writers didn't want studios to use their work to teach chatbots how to write, feeding AI old scripts so that chatbots could generate writing in a similar style. The writers also worried that studios would ask

chatbots to rewrite or refine the first drafts of their work for scenes or whole shows. That's the nightmare scenario, said John August, who is on the Writers Guild negotiating committee. The studios had initially said that too much was unknown about the technology and that the Guild would need to wait to discuss it in future contract negotiations. But over the weekend, the studios proposed a few paragraphs to be inserted into the new contract that addressed a writer's concern about AI and old scripts. The two sides spent several hours negotiating the language on the final night of talks. A separate actor's strike led by SAG AFRA is continuing. The actors want 2% of the total revenue generated by streaming shows, something that studios have said is a non-starter.

There are no talks scheduled between those two sides, end quote. Finally, this is an interesting data point that maybe we would have been able to intuit just from what we cover on this show day in and day out. According to Counterpoint Research, the number of smartphone brands fell from more than 720 in 2017 to almost 250 smartphone brands in 2023. Apparently, a lot of local brands have been losing market share due to a maturing user base, the big 5G shift, and more, but basically this is just the case of the big getting bigger. Quote, a maturing user base, improving device quality, longer replacement cycles, economic headwinds, supply chain bottlenecks, and major technological transitions such as 4G to 5G have gradually whittled

down the number of active brands from their volumes over the years. For example, local smartphone brands once known as local kings like Micromax in India and Symphony in Bangladesh, have lost significant share or even exited over the last five years. Strikingly, the decline in the number of active brands is coming largely from local brands, while the number of global brands has remained mostly consistent. Most local brands operate in lower price bands and in regions that have fragmented markets across wide geographies like Asia Pacific, Latin America, and Middle East and Africa. In a rapidly evolving smartphone industry, small brands have struggled to keep up with big brands across many fronts. While big brands have continued to invest in R&D, manufacturing,

and capacity building, small brands have been largely dependent on white label devices. Furthermore, more large promotional and marketing events and big-name brand ambassador tie-ups from sports and movies are commonplace for big brands, which most small brands don't have the resources to do. Small brands capitalized on the markets transition from 2G to 3G 4G, benefiting from strong entry-tier demand, particularly in Africa, Asia, and Latin America. However, the needs of the average mobile phone consumer have been evolving since, and the user base has matured. Therefore, there is now a greater demand for better specifications than design,

brand value and ecosystem integration. The rise of Chinese brands like Xiaomi, Apo, and Vivo has also accelerated the decline of small brands. Chinese brands have been able to introduce significantly better smartphones at aggressive price points providing customers better value for their money, end quote.

So I want to finally, officially take the wraps off my little AI experiment that you've been hearing about all summer. You can find it live at resume writing.com. Basically, we've just, ginned up a completely modern AI resume product. If you sign up and train the system on what we call your foundation resume, which is your career history, then what you can do is subscribe for AI

credits that allow you to refine that resume over and over again with the AI. Basically, you can just cut and paste in a job opening that you find online and it will tweak your resume to target the keywords in that job opening. It kind of rewrites the whole thing. So basically every job you apply to in a job search can now have a specific resume targeted to that exact job as opposed to just, you know, one size fits all. When Chris and I talk about on the show finding people in the space that are using AI to create things that weren't possible before, this is in a somewhat modest way what we're talking about. Before now, you either had to use a one size fits all resume for every job you apply to or every time you applied, you'd have to

rewrite the resume from scratch, but the AI can do that for you. This product basically automates that whole process and allows you to do it at job search scale. One resume for every different job you apply to. Check it out at Resumaywriting.com, because we're still in beta, so all feedback is welcome right now. And also spread the word about it so we can get more beta testers as customers and refine the product more. I'm going to do a bonus episode with UpTech. They're the development team that put this together for. me. I've learned so much about what is and what is not possible with AI at the moment. So we're going to have a conversation and walk through how we constructed this, but also what

we've learned from doing that. Check out uptech at uptech.tec.tec. I could not be more complimentary about what they do. They basically are a full-stacked product studio and software development company. There's nothing that you might want to do that they can't help you out with from apps to websites to full product development. They were amazing to work with. Remember, I had hit the wall, essentially, and they completely helped us jump over that wall. They were completely responsive and almost intuitive,

like they anticipated UI and UX things before I could even discover that we needed them. Could not recommend them more highly. Uptech.com. Pretty much now officially the development partner of this podcast, of the mutant podcast army. Uptec.com. Talk to you tomorrow.