Tech Brew Ride Home - Fri. 02/24 – Record Scratch On The Adobe/Figma Deal

Episode Date: February 24, 2023

Is the DOJ about to sue to stop of Adobe’s acquisition of Figma? Confirmation of the lower end mixed reality headset from Apple due next year. A reporter used an AI-generated replica of a voice to bypass the voice verification system of a major bank. And, of course, the Weekend Longreads Suggestions. Sponsors: HelloFresh.com/techmeme65 and code techmeme65 for 65% off Links: DOJ Preps Antitrust Suit to Block Adobe’s $20 Billion Figma Deal (Bloomberg) Kuo: Apple to Launch High-End and Low-End Versions of Second-Generation Headset in 2025 (MacRumors) How I Broke Into a Bank Account With an AI-Generated Voice (Motherboard) Weekend Longreads Suggestions: What Is ChatGPT Doing … and Why Does It Work? (Stephen Wolfram) Getting My ChatGPT Plus Subscription Is an Inflection Point (HackerNoon) Why China Didn’t Invent ChatGPT (NYTimes) Meet the $10,000 Nvidia chip powering the race for A.I. (CNBC) Inside Meta’s Push to Solve the Noisy Office (WSJ) Are bioinspired drones the next big thing in unmanned flight? (TNW) 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 Friday, February 24th, 2020. I'm Brian McCullough. Today is the DOJ about to sue to stop Adobe's acquisition of Figma.

Confirmation of the lower-end mixed reality headset from Apple due next year. A reporter used an AI-generated replica of a voice to bypass the voice verification system of a major bank. And, of course, the week on long-read suggestions. Here's what you miss today in the world of tech. Record scratch. A source has told Blue, Bloomberg that the DOJ is preparing to file an antitrust lawsuit to block Adobe's $20 billion Figma acquisition as soon as March. Adobe expects to close the deal in 2023, but now,

quoting Bloomberg. A case is expected to be filed as soon as next month, although the timing could slip, said one of the people, all of whom asked for anonymity to discuss the confidential probe. The deal needs approval from several antitrust authorities, and the merger agreement allows for a possible extended regulatory review with an outside completion deadline of March 24. Adobe had a meeting with the DOJ yesterday, according to another person. The deal also faces an antitrust review in the European Union after the Block's antitrust watchdog said it had received requests from national regulators to look into the deal. The UK Competition and Markets Authority is reviewing the merger as well, and the three jurisdictions

often coordinate on their investigations. Adobe, the dominant force for years in software such as Photoshop and Illustrator for design professionals announced the deal to acquire Figma in September. The purchase is a massive bet that more creative work will be done by small businesses and everyday users on the web, a market that Figma has rapidly seized. While Adobe has introduced less expensive streamlined products for that audience, most of its offerings are still heavyweight programs aimed at specialists. The company still expects to close the transaction in 23 and is engaged in, quote, constructive and cooperative discussions with regulators in the U.S., UK, and EU, among others, according to an Adobe spokesperson, the Justice Department declined to comment.

Adobe tried to buy Figma in 2020 and 2021 as the startup rapidly gained steam, according to a filing with details on how the merger came together. Eventually in 2022, Figma accepted and offered double its valuation at a time when many of its peers were seeing decreases. Wall Street Analyst saw the price tag as revealing severe competitive pressures on Adobe. Figma unsuccessfully solicited a bid from Microsoft 2 before accepting Adobe's offer. The lawsuit represents Adobe's most serious spat with antitrust regulators since the 1990s when it was forced to divest a program that competed with its illustrator graphic design program as part of a merger with Aldus Corporation. It saw another Justice Department investigation in 2005 over its acquisition of Macromedia's suite of web development

software, but the deal was ultimately allowed to close without a legal challenge, end quote. Confirmation from our buddy Ming Chi Kuo that Apple plans a cheaper second-gen AR MR headset with two high-end and two low-end models built by Lux Kaysicht and Foxcon, respectively, likely coming in 2025, quoting Mac rumors. That report said Apple's first headset will be extremely expensive, with industry estimates ranging from $3,000 to $5,000, while Apple's second-generation headset will have a more affordable price within the territory of a high-end Mac computer. Now it seems Apple is planning a two-tiered series of the second-generation device to appeal to a wider customer base, similar to the way it offers both standard and more premium pro-branded iPhones each year.

In his latest report, Quo says that Pegatron is gradually withdrawing from Apple's headset business and will likely transfer its AR, MR, development team, and production resources to Lux KSICT, a joint venture between Lux Share and Pegatron, led by LuxShare ICT in the first half of 2023. This will see LuxShare ICT taking over the subsequent design and production of the high-end version of the second-generation headset. Such changes will lead to, quote,

the subsequent acceleration of reducing the cost of the headset, which is what Apple expects, adds quo. Apple's plans to release a cheaper version, version of its ARVR headset were first reported last month by the information's Wayne Ma and Bloomberg's Mark Gurman. The headset would supposedly use more affordable components, such as lower-resolution lenses. According to the information's significant work on the second-generation device started last year, and at the time, the goal was to launch the cheap headset in 2024. Bloomberg previously reported that Apple's budget mixed reality headset could arrive in either 2024 or 2025, end quote.

A reporter says he used an AI-generated replica of a voice to bypass the voice verification system of the UK's Lloyd's Bank, and then he subsequently accessed an account's information. Given this topic, given that we've been experimenting with just this thing, I can't resist doing the rest of this segment in my own robot voice, and I've been using 11 labs, the very tool that will be mentioned in the story. I've continued to train my robot voice, so see if it's improved to your ears, quoting motherboard. On Wednesday, I phoned my bank's automated service line. To start, the bank asked me to say, in my own words, why I was calling. Rather than speak out loud, I clicked a file on my nearby laptop to play a sound clip. Check my balance, my voice said. But this wasn't actually my voice. It was a synthetic clone I had made using readily available artificial intelligence technology. Okay, the bank replied. It then asked me to enter, or say my date of birth, as the first piece of authentication. After typing that in, the bank said, please say, my voice is my password.

Again, I played a sound file from my computer. My voice is my password, the voice said. The bank's security system spent a few seconds authenticating the voice. Thank you, the bank said. I was in. I couldn't believe it. It had worked. I had used an AI-powered replica of a voice to break into a bank account. After that, I had access to the account information, including balances, and a list of recent transactions and transfers. Banks across the U.S. and Europe use this sort of voice verification to let customers log into their account over the phone. Some banks tout voice identification as equivalent to a fingerprint, a secure and convenient way for users to interact with their bank. But this experiment shatters the idea that voice-based biometric security provides foolproof protection in a world where anyone can now generate synthetic voices for cheap or sometimes at no cost.

I used a free voice creation service from 11 Labs, an AI voice company. Now, abuse of AI voices can extend to fraud and hacking. Some experts I spoke to after doing this experiment are now calling for banks to ditch voice authentication altogether, although real-world abuse at this time could be rare. For this particular attack, a fraudster would also need the target's date of birth. But thanks to a plethora of data breaches, brokers, or people sharing personal details online, a date of birth is often readily available. Over the last few weeks, I have tested a few AI voice generation services. Most of them had problems or limitations with recreating my British accent, which would be necessary to access the Lloyd's bank account. Eventually, I used 11 labs, which handled the accent well. To create the voice, I recorded about five minutes of speech and uploaded it to 11 labs. For the audio clips, I read sections of Europe's data protection law.

A short while later, the synthetic voice was ready to use, with it saying whatever text was entered into 11 Labs' site. The experiment of entering the bank account failed multiple times with Lloyd Bank's system saying it could not authenticate the voice. After making some tweaks on 11 labs, such as having it read a longer body of text to make cadences sound more natural, the generated audio successfully bypassed the bank's security, end quote. Time for the week on Lurie's suggestions. It's AI heavy again this week. But if you want the most in-depth explanation of what chat GPT is actually doing, how and what why it works the way it does. Stephen Wolferham has you covered with something like 50,000 words

on the whole technical underpinnings of this stuff. Again, it's super long, super detailed, but it's filled with a lot of aha insights like these. Quote, so how does neural net training actually work? Essentially what we're always trying to do is find weights that make the neural net successfully reproduce the examples we've given. And then we're relying on the neural net to interpolate or generalize between these examples in a reasonable way, end quote. And quoting again, the fact that there's randomness here means that if we use the same prompt multiple times, we're likely to get different essays each time. And in keeping with the idea of voodoo, there's a particular so-called temperature parameter that determines how often lower-ranked words

will be used. And for essay generation, it turns out that a temperature of 0.8 seems best. It's worth emphasizing that there's no theory being used here. It's just a matter of what's been found to work in practice. And for example, the concept of temperature is there because exponential distributions familiar from statistical physics happen to be being used, but there's no physical connection, at least so far as we know, end quote. And this post on Hacker Noon was interesting to me. It's one developer sharing his experience using generative AI to code for the first time. Quote, There are some moments in the life of a geek like me that are truly inflection points, like the first time back in 1993, that I saw the web browser mosaic on the computer at the university I was working at. Just like in the movies, I started daydreaming of all the possibilities that the web would open up.

I just got my chat GPT plus subscription today. They finally opened up here in Europe. In the previous days, it was only available in the U.S. My chat GPT moment was very similar to the Mosaic Web Browser 1. I've been using it for a few weeks now, and I feel just like I felt back then that this is going to change things in a big way, end quote. The New York Times gets it something that we discussed this very week. Why didn't China invent ChatGPT? A lot of people say that China is ahead of the West when it comes to AI. Well, maybe it functionally can't create something like ChatGPT. The government has been the biggest barrier to AI, its obsession with censorship, perhaps its heaviest club.

The availability of a wide range of data is crucial to developing technology like chat GPT, and that is increasingly harder to come by in a censored online environment. Today, jokes circulate that capture the dark mood among tech people. A popular one? We need to teach machines not only how to speak, but also how not to speak. Beijing has punished companies sometimes severely to enforce its censorship protocols. Duo Lingo, which is in the seemingly non-controversial business of teaching people new languages, was taken out of Chinese app stores for nearly a year to, quote, enhance its content regulation,

according to Chinese media reports. Many of us in the internet industry are faced with two problems when making a product. Either our products don't involve speech, or they have to undergo a lot of censorship, said how Piquang, a former entrepreneur and programmer in the northern city of Tianjin. Big companies can't afford it, but smaller companies can't, he said. If small companies can't do this, it stifles innovation, end quote. You might have heard that Nvidia's market cap has gone up by something like $100 billion in the last month alone. I think the stock was up something crazy, like 15% in a single day at one point this week. And you want to know why?

The type of chips that Nvidia specializes in are, once again, perfectly positioned to be used for this AI revolution. Quoting CNBC, powering many of these new AI applications is a roughly $10,000 chip that's become one of the most critical tools in the artificial intelligence industry, the Nvidia A-100. The A-100 has become the workhorse for artificial intelligence professionals at the moment, said Nathan Benyatch, an investor who publishes a newsletter and report covering the AI industry, including a partial list of supercomputers using A-100s. Invidia takes 95% of the market for graphics processors that can be used for machine learning, according to New Street research. The A-100 is ideally suited for the kind of machine learning models that power tools like chat GPT, Bing, AI, or stable diffusion.

It's able to perform many simple calculations simultaneously, which is important for training and using neural network models. The technology behind the A-100 was initially used to render sophisticated 3D graphics and games. It's often called a graphics processor or GPU, but these days, Nvidia's A-100 is configured and targeted at machine learning tasks and runs in data centers, not inside glowing gaming PCs. Big companies or startups working on software like chatbots and image generators require hundreds or thousands of Nvidia's chips and either purchase them on their own or secure access to the computers from a cloud provider. Hundreds of GPUs are required to train artificial intelligence

models like large language models. The chips need to be powerful enough to crunch terabytes of data quickly to recognize patterns. After that, GPUs like the A100 are also needed for inference or using the model to generate text, make predictions, or identify objects inside photos. This means that AI companies need access to a lot of A100s. Some entrepreneurs in the space even see the number of A-100s they have access to as a sign of progress, end quote. Then in non-AI news from the Wall Street Journal, has meta-reinvented the office cubicle. No, I'm not talking about meetings in the metaverse. I'm talking about actual cubicles for use in meat space.

Quote, it's a cubicle. That is, a noise-canceling cubicle designed using some of the same principles found in a soundproof, echo-free, and echoic chamber. The Cube, which the company is beginning to roll out to offices worldwide after months of development, absorbs sound from multiple directions, says John Tenanis, Vice President of Global Real Estate and facilities at Meta. It's like a self-cocoon, he said, end quote. And finally today from the next web is the next big thing in drones, actually little things, tiny drones, drones, not only the size of a bird or a bug or a dragonfly, but also drones that function like them too, flapping around in the breeze.

Bio-inspired drones, quote, we understand there are things within nature that have developed really excellent solutions to problems that we also face as humans, says Ian Foster, head of engineering at Animal Dynamics, who is one of the 91 staff members at the firm. That's something Matej Kyrasek sees echoes of in his own company. Kerasek is the founder of Dutch startup Flapper, previously Flapper Drones, a spinoff from Delft University of Technology, which counts to employees and has raised $100,000 euros in seed money. The University Project had been going for the best part of two decades

and was designed to try and develop a bio-inspired drone that was lightweight. The size and scale aspect is a necessary evil, Kerasex says. One of the key advantages of bio-inspired drones is that they need to be small because of the physics, he explains. That enables them to do more detailed, fine-skilled tasks that bigger drones can't, making a virtue out of what may initially seemed like a limitation. Quadcopters are sensitive to damage, he says. If they hit something, they break. Flapping ones might be less sensitive and potentially something that could restart again, even if it crashes, end quote. Animal Dynamics's Stork drone doesn't see size as an issue either. Its parafoil can carry a payload of 135 kilograms up to 400 kilometers, thanks to the nature-inspired

revolution of simply gliding for kilometers without powering the engine. Something that Foster, the company's head of engineering, believes, makes it useful for work in less built-up areas. We want to be able to operate in places where they're very remote, he says. We're delivering aid to an area that has collapsed infrastructure. There isn't going to be an airport there, end quote. But for drone companies like Flapper, trying to find a niche in more built-up populated environment, safety is one area where it sees its range of bio-inspired drones as offering a key point of differentiation.

If you fly into something with a conventional drone, the sharp propellers could cut into things. But with soft wings, they actually. actually bounce off objects, Kerasex says, end quote. Guess what? After a long hiatus, we have a bonus episode this weekend, a portfolio profile episode. What if I told you I found a company that has found a true application for VR? No, not virtual office meetings. No, not games, but actual science. Actually building molecules. And the founders were longtime listeners to the show. That's why they approached

the ride home fund initially, so listen to fellow members of the mutant podcast army that are building something really, really cool. We'll drop that episode tomorrow. Talk to you on Monday.