Tech Brew Ride Home - Fri. 10/04 – AI Slop Comes To Podcasting

Episode Date: October 4, 2024

Google has updated Lens and is taking a page out of Perplexity’s book. OpenAI’s new canvas workspace. Why an upcoming iPhone SE might have some interesting internals. And, of course, the Weekend Longreads Suggestions. Links: Google Lens now lets you search with video (The Verge) Google brings ads to AI Overviews as it expands AI’s role in search (TechCrunch) OpenAI launches new ‘Canvas’ ChatGPT interface tailored to writing and coding projects (TechCrunch) iPhone SE 4 to feature Apple’s first 5G modem, A18 chip, same cameras as iPhone 15 (9to5Mac) 159 employees are leaving Automattic as CEO’s fight with WP Engine escalates (TechCrunch) Weekend Longreads Suggestions: Saving Cyberpunk 2077: How CD Projekt Red recovered from one of video games' most disastrous launches (EuroGamer) The Flying Car Is Finally Here. It’s Slightly Illegal. (Intelligencer) 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, October 4th, 2024. I'm Brian McCullough today. Google has updated Lens and is taking a page out of perplexity's book, Open AI's new canvas workspace, why an upcoming iPhone SE might have

some interesting internals, and of course the weekend long-read suggestions. Here's what you miss today in the world of tech. Google has updated Google Lens to let users take a video and use their voice to search that, showing an AI overview and search results, rolling out to Search Labs on mobile, quoting the verge. If you can't capture what you want to search for with just a picture, Google Lens will now let you take a video and even use your voice to ask about what you're seeing. The feature will surface an AI overview and search results based on the video's contents and your question. It's rolling out in Search Labs on Android and iOS today. Google first previewed using video to search at I.O. in May, as an example, Google says

someone curious about the fish they're seeing at an aquarium can hold up their first. phone to the exhibit, open the Google Lens app, and then hold down the shutter button. Once Lens starts recording, they can say their question. Why are they swimming together? Google Lens then uses the Gemini AI model to provide a response similar to what you see in the GIF below. When speaking about the tech behind the feature, Rajan Patel, the vice president of engineering at Google, told Diverge that Google is capturing the video as a series of image frames, and then applying the same computer vision techniques previously used in lens. But Google is taking things a step further by passing the information to a custom Gemini

model developed to, quote, understand multiple frames and sequence, and then provide a response that is rooted in the web, end quote. There isn't support for identifying the sounds in a video just yet, like if you're trying to identify a bird you're hearing, but Patel says that's something Google has been, quote, experimenting with, end quote. Google also rolled out ads in AI overviews for when a user's question has a commercial angle and identifies the ads with a sponsored header on mobile in the US. So when you get those AI overviews now on Google search, expect there to be ads soon. And more than that, Google is also now doing flat out AI organized search results in the US on mobile, showing users AI aggregated pages of content from forums,

articles, YouTube videos, and more. So doing a perplexity. I, suppose. Quoting TechCrunch. Searches about recipes and meal inspiration, for example, what are some good vegetarian appetizers or dinner ideas that wow may pull up an AI aggregated page of content from around the web, including forums, articles, and YouTube videos. A customized Gemini model generates a full page experience with results that are relevant and organized, Bell explained, referring to Google's Gemini family of AI models. With these AI organized results pages were surfacing more diverse content formats from a more diverse content set, end quote. Google says it plans to expand these pages to other categories of searches in the coming months.

Publishers may be the collateral damage, however. One study found that AI overviews could negatively affect about 25% of publisher traffic due to the de-emphasis of web page links. On the revenue side, an expert cited by the New York Post estimated that AI-generated overviews could lead to more than $2 billion in publisher losses, thanks to the resultant decline in views. AI-generated search results from Google and rivals don't appear to be cratering big publishers' traffic yet. In their most recent earnings, Ziff Davis and Dot Dash Meredith Parent IAC characterized the impacts as negligible. But that may change as Google, which commands over 81% of the global search market expands AI overviews and AI organized pages to more users and queries.

According to one estimate, AI overviews were only showing for about 7% of searches in July as Google dialed back the feature to make adjustments, end quote. OpenAI has launched Canvas, a chat GPT interface with a workspace for writing and coding projects into beta for Plus and Team users. This is similar to Anthropics artifacts feature, apparently, coding TechCrunch. AI chatbots today can't complete large projects from a single prompt, but they can often create a good starting point. Editable workspaces like Canvas allow users to fix parts of an AI chatbot's output that are wrong without having to scrutinize their prompt and generate a whole new stretch of code.

The product opens a separate window beside the normal chat window with a workspace for writing and coding projects. Users can generate writing or code directly in the canvas, then highlight sections of the work to have the model edit. Canvas is rolling out in beta to chat GPT Plus and teams users on Thursday and Enterprise and EDU users next week. Several consumer AI providers are converging around editable workspaces as a practical way to use generative AI.

ChatGPT's new interface offers similar features to Anthropics artifacts launched in June and the viral coding companion cursor. OpenAI is racing to match competitor offerings and launch entirely new capabilities in ChatGBT as a means to grow its paid user base. This is just a more natural interface for collaborating with ChatGBTGBT, said OpenAI product manager Daniel Levine in a demo with TechCrunch. In our demo, Levine had to select GPT40 with Canvas from ChatGPT's model picker. drop-down window. However, OpenAI says Canvas windows will just pop out when ChatGPT detects a separate workspace could be helpful, say for longer outputs or complex coding tasks.

You can also write, use Canvas to automatically open a project window. Levine showed TechCrunch how ChatGPT's new features could help write an email. Users can prompt ChatGPT to generate an email, which will then pop out in the Canvas window. Then users can toggle a slider to adjust the length of the writing to be shorter or longer. You can also highlight speech. specific sentences and ask ChatGBTGBT to make changes such as make this sound friendlier or add emojis. Users can also ask ChatGPT to rewrite the whole email as is in another language. The features for the coding Canvas are slightly different, however. Levin prompted ChatGBTBT BT to create an API web server in Python which spawned in the canvas window. By pressing an

add comments button, ChatGPT will add inline documentation to explain the code in plain English. Further, if you highlight a section of code that ChatGPT created, you can ask the chatbot to explain it to you or ask questions about it. ChatGPT is also getting a new review code button which will suggest specific edits for the code in the window, whether generated or user written for them to approve, edit themselves, or decline. If they press approve, ChatGPT will take a stab at fixing the bugs itself, end quote. There continues to be a bunch of rumor that Apple is going to hold some sort of October event, where they will likely refresh the Mac lineup with M4 chips, maybe release a new iPad Mini. But there are also rumors about the return of the iPhone SE, an iPhone SE4.

The Scuttlebutt says that probably won't come until spring of next year, but here's the speculation about what that device might have, quoting 9 to 5 Mac. While the current generation iPhone SE looks like an iPhone 8, the new model will have a more modern look similar to an iPhone 14. This includes flat sides and an OLED panel with a notch at the top. Our sources confirm that the iPhone SE4 codenamed V-59 will have the same 1170 by 2532 display resolution as the 6.1-inch iPhone 14. Apple will also bring face ID to the iPhone SE4, putting an end to the home button with touch ID on the iPhone. However, the dynamic island will remain exclusive to the more premium models. Previous reports suggest that Apple wants to make

the new SE compatible with Apple intelligence features, and our sources say that this will be possible thanks to the A18 chip with 8 gigabytes of RAM. Apple will use the same variant of the latest system on a chip used in the base iPhone 16 with a five-core GPU. When it comes to the camera, the iPhone SE4 will have the same 48-mixel wide camera and 12-mapsule front camera as the iPhone 15 and 15 plus. However, we've heard that the new SE won't have ultra-wide or telephoto lenses. Apple will likely take advantage of the 48 megapixel sensor to say that the new SE is capable of taking photos with 2X optical zoom. But surely one of the main highlights of the iPhone SE4 will be Apple's first 5G modem. The company acquired Intel's modem division in 2019 in an attempt to dispose of Qualcomm's modems,

but multiple reports suggest that the prototype modems developed by Apple have suffered multiple failures. It seems that Apple is confident of finally getting its first 5G modems into the hands of a lot of people, 9 to 5 Mac's sources say that the V-59 is equipped with a wireless modem designed by Apple code named Centauri. The modem is quite ambitious and will also handle Wi-Fi, Bluetooth, and GPS. While Apple will certainly benefit in terms of money from having its own modem, the company will also use it to improve the integration between hardware and software. 9-to-5 Mac has learned that the new modem will drastically reduce battery consumption, especially when users switch on low-power mode on the iPhone, end quote.

This whole story continues to be really weird. Now, Matt Mullenweg says 159 automatic employees are around 8.4% of staff have taken a, quote, generous severance package offered to those who disagreed with WordPress's direction in recent weeks, quoting TechCrunch. In a blog post, Malinweg said the package offered $30,000 or six months of salary, whichever is higher, but the employees who took it would not be eligible to be rehired by automatic. Nearly 80% of people who took the offer worked in the company's ecosystem-slash WordPress division, and the rest were in Automatic's Cosmos business consisting of apps like

Pocketcast, Day 1, Tumblr, and CloudUp. Mullenweg, who co-created WordPress and is arguably the face of the open source project tried to put a positive spin on the announcement, writing that the company, quote, decided to design the most generous buyout package possible. We called it an alignment offer. HR added some extra details to sweeten the deal. we wanted to make it as enticing as possible, he wrote and later on added, 159 people took the offer, 8.4% of the company. The other 91.6% gave up $126 million of potential severance to stay, end quote. Okay.

Time for the weekend long read suggestions. First up, remember when Cyberpunk 277 launched and everybody hated it? Well, it turned out the promised improvements and fixes that Game Maker CD Project read made to the game have actually worked and people generally think the game is good now. So Eurogame took a look at how the company turned things around. Quote, Naturally, plenty wonder why the studio didn't simply delay it again for longer to properly allow for things together. For one, as Noakowski says, the breakdown in communication and outsized confidence meant the decision makers simply felt they didn't need to. The second was

sheer exhaustion. The team was really, really tired, Noah Kowski says, could almost feel that people just wanted some sort of closure. If we pushed it further, probably, or at least that's how we felt, that would mean a lot of challenges for the team. It turned out that because we didn't, it was probably even harder. We figured A, it's going to be probably fine, and B, it could actually mean more damage for the team if we're going to keep going. And we didn't realize we were in so much trouble, I think, most importantly. Because if the first part was clearly visible, I don't think we would have taken that decision despite number one and number two. Nobody sane does something like that to a game that's been in development for so long

that we were betting on so hard, end quote. Altogether, it made for a damaging cocktail that couldn't be unstirred. Ambitious plans for a seamless open world with no loading tricks, an initially simultaneous release across nine platforms, old generation consoles, pro editions, new generations, PC, and Stadia, visual fidelity to push the limits of high-end PCs, a series of new systemic mechanics alongside a new IP, all combined with an approach where the team built its own engine as it went, almost tripled in size, and moved to emergency remote work at the height of the pandemic. Throughout all this, CG Project continued to hold faith in the methods that had seen it through past launches that went down to the wire.

Michael Nowakowski described it to me best in labeling the studio's unshakable optimism as simply magical thinking. Eventually, CD Project Red had at last realized what the problems were. Exhausted and crestfallen the team took time to reset over. Christmas and New Year. Then, at the beginning of 2021, those who didn't leave the company returned to work and put the same unshakable faith that partially got them into this mess to another use, digging themselves back out, end quote. And finally, and this is not a joke, actual practical flying cars are here, and they're almost legal, like almost. Quoting

intelligents are, most mornings when the air lies still on the ridges of the North Cascades in central Washington State, Tim Lum climbs into his personal flying car, a 14-foot-long bean-shaped craft called A Black Fly, straps himself in, and sets the machines for rotor blades whirring. As the 61-year-old retired smokejumper levitates into the sky, the landscape opens up below him, the forest stretching along the ridge, and the farmland sprawling across the valley floor below. The aircraft swings forward into horizontal flight, and Lum zips off, flowing along the contours of land taking in the scenery. It's stunning, very dramatic, he tells me, cliffs and trees and valleys. I've been writing paragraphs like that for a decade for magazines like popular mechanics and popular

science, imagining a time in the not-so-distant future when the long-awaited promise of flying cars, more officially known as electric vertical takeoff and landing vehicles, or EvTol, is finally made a reality. This time, though, the scene is not a flight of fancy. LUM is a real person, and he really does fly a personal flying machine, typically, around five times a day. If you have $190,000 on hand, you too could buy one, or if your budget is more modest, you can book a rental ride and a different kind of electric flying vehicle for $249. At long last, the era of the flying car is here, but there is a catch. The EVTOLs that are currently available are not, strictly speaking, legal. The entire fledgling industry, such as it is,

exists in a kind of shadowland where it's unclear what exactly the rules are and what will happen if you break them. For the manufacturers, it's a gamble, the kind of regulatory arbitrage that could allow them to jump ahead of more careful rule-following competitors and become an industry dominating colossus like Airbnb or Uber, or could devolve into lawsuits and enforcement actions. Move fast and break things is the Silicon Valley way, but it's very much the opposite of the safety-first mindset of the rest of the aviation industry, where if things crash and burn, they do so literally, and the public and government are not quick to forgive, end quote. It's only been a week, but content from that notebook LM stuff is everywhere.

Just do a cursory search on YouTube for almost any topic. And you'll find a bunch of, in quotes, podcasts that have popped up all of a sudden to people talking about a topic. And same if you search podcasts themselves on Spotify or whatever. There's a whole bunch of new podcasts up there. I guess the slop has finally come to podcasting too. Awesome. Will this put a premium on actual human voices? and thoughts, or no, hard to say.

If the last 20 years has taught us anything, it's that the scale of spam and slap can overwhelm almost anything, but then again, maybe people will want the real thing. I don't know. As Nilai Patel said on threads, every platform company is about to be at war with itself as the algorithmic recommendation AI team

tries to fight off the content made by the generative AI team. Yay. Anyway, no bonus. episodes this weekend. Talk to you on Monday.