Tech Brew Ride Home - Wed. 06/14 – EU Threatens Google With A Breakup

Episode Date: June 14, 2023

The European Commission might have just taken the first steps toward separating Google from its cash cow. Reddit’s CEO keeps saying not so soothing things. New Open AI and Meta releases for… AI stuff. Cybersecurity premiums are skyrocketing. Spotify has fully shifted its podcast strategy. And a look at the tech job market as AI takes over. Sponsors: NewtonX.com/techmeme CrashPlan.com Links: EU suggests breaking up Google’s ad business in preliminary antitrust ruling (The Verge) Reddit CEO tells employees that subreddit blackout ‘will pass' (The Verge) Google is getting a lot worse because of the Reddit blackouts (The Verge) OpenAI intros new generative text features while reducing pricing (TechCrunch) Cyber Insurance Premiums Surge by 50% as Ransomware Attacks Increase (Bloomberg) Spotify Takes a Sharp Turn With Its $1 Billion Podcast Division (Bloomberg) Meta AI researchers unveil I-JEPA, a computer vision model that learns more like humans do (SiliconAngle) Layoffs and AI Are Changing Tech’s Once-Invincible Job Market (Bloomberg) 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 Meme right home for Wednesday, June 14th, 2023. I'm Brian McCullough today. The European Commission might have just taken the first steps towards separating Google from its cash cow. Reddit's CEO keeps saying not so soothing things.

New Open AI and meta releases for AI stuff. Cybersecurity premiums are skyrocketing. Spotify has fully shifted its podcast strategy and a look at the tech job market as AI takes over. Here's what you miss today in the world of tech. The European Commission has issued a preliminary anti-trust complaint against Google over its ad business, suggesting breaking up the unit. Google can now reply, but it's pretty much the nuclear scenario they've been fearing all along at this point. Quoting the verge, in a preliminary opinion, the regulator says Google has abused its dominant position in the digital advertising market. It says that forcing Google to sell off parts of its business may be the only remedy if the company is found guilty of the charges.

This would be a significant move targeting the main source of the search giant's revenue and a rare example of the EU recommending divestiture at this stage in an investigation. The commission has already fined Google over three prior antitrust cases, but has only previously imposed behavioral remedies changes to its business practices. The statement of objections issued today is an important step in the EU's investigation but does not prejudge its outcome. Google will now have the opportunity to reply in writing and request a hearing, after which the commission will, decide whether Google has broken antitrust law in the block. If found guilty, the EU's competition regulator can also find Google up to 10% of its global sales and impose various changes to its business. In a statement, Google's vice president of global ads, Dan Taylor said the company disagrees with the commission's position and called digital

advertising a, quote, highly competitive sector, end quote. Reddit update for you, over 300 subreddits, including R-slash-a, R-slash music, and R-slash video. say they plan to extend their protests indefinitely. More than 8,400 subreddits have now gone private or into a restricted mode, but a lot of those are supposed to turn back on today. Now, this should help calm the waters a bit. In a staff memo, Reddit's CEO apparently said, quote, like all blowups on Reddit, this one will pass, end quote. In case you missed it, I'm being facetious. That probably won't calm the waters at all, quoting the verge. Huffman says the blackout hasn't had, quote, significant revenue impact and that the company anticipates that many of the subreddits will come back online by Wednesday.

There is a lot of noise with this one. Among the noisies we've seen, please know that our teams are on it. And like all blowups on Reddit, this one will pass as well. The memo reads. We absolutely must ship what we said we would. The only long-term solution is improving our product. And in the short term, we have a few upcoming critical mod tool launches we need to nail, end quote. Huffman also warns employees about wearing Reddit items in public, saying the anger directed at the API pricing change could make them, quote, the object of users' frustrations, end quote. Even if your favorite subreddit hasn't gone dark, this still could be affecting you. You know that thing where when you search on Google, you add Reddit to the search to get better results. Well, that's being impacted, quoting the verge again. And even if you don't rely on the Reddit trick like I do, Reddit links often show up at the top of search results. way, meaning that many people who don't regularly use the platform have probably found some

useful information on the site. In Tears of the Kingdom, for example, I'm currently on the hunt for a specific piece of armor, but when I searched for tips about it last night, I found that R-S-T-K is private. I've been having some issues with a pocket door in my home, but I can't read a promising R-S-D-Y post because I can't access the community. I've been meaning to find some new good music, but R-slash-Mu-Mu-S-Mu-S-S-S-S-S-S-S-S-S-S-S-Wy-W-D-Dark. Sure, Google can provide me answers for any one of those needs. Other sites have great guides for Tears of the Kingdom, but none of those have the conversational and community elements that makes Reddit so dang useful. I like perusing the comments below a post to see other recommendations, points of view, and other

links to relevant sources, and then seeing other people discuss the merits of those additions to the thread, end quote. You're speaking to my specific concerns, Jay Peters at the verge. OpenAI has released new GPT3.5 Turbo and GPT4 versions that include a new function calling feature. They've also reduced the pricing of the original GPT3.3.5 Turbo by 25%. Quoting TechCrunch. As OpenAI explains in a blog post, function calling allows developers to describe programming functions to GPT3.5 Turbo and GPT4 and have the models create code to execute those functions. For example, function calling can help to create chatbots that answer questions by calling external tools,

convert natural language into database queries, and extract structured data from text. These models have been fine-tuned to both detect when a function needs to be called and to respond with responses that adhered to the function signature. OpenAI writes, function calling allows developers to more reliably get structured data back from the model, end quote. Beyond function calling, OpenAI is introducing a flavor of GBT 3.5 turbo with a greatly expanded context window. Context window measured in tokens or raw bits of text refers to the text the model considers before generating any additional text. Models with small context windows tend to forget the content of even very recent conversations, leading them

to veer off topic, often in problematic ways. The new GPT3.5 turbo offers four times the context length, 16,000 tokens of the vanilla GPT3.5 turbo at twice the price. Three-tenths of a cent, per 1,000 token inputs, i.e. tokens fed into the model, and four-tenths of a penny per 1,000 output tokens, tokens the model generates. OpenAI says that it can ingest around 20 pages of text in a single go, short of the hundreds of pages that AI startup Anthropics flagship model can process notably. OpenAI is testing a version of GPT4 with a 32,000 token context window, but only in limited release. On the plus side, OpenAI says that it's reducing pricing for GPT3.5 turbo, the original, not the version with the expanded context window, by 25%. Developers can now use the model for

0.0015 per 1,000 token inputs and 0.002 per 1,000 output tokens, which equates to roughly 700 pages per dollar, end quote. According to a study, U.S. cyber insurance premiums surged 50% last year. Due to ransomware attacks and online commerce, as policyholders claimed $7.2 billion, in damages, which is up three acts in just three years. Quoting Bloomberg. Systemic risk is an ongoing concern, Fred Eslami, an AM Best Associate Director said in a statement, ultimately the coverage provided to insureds may be decided by the risk

appetite of the insurer and to a certain extent, the coverage that reinsurers are willing to provide, end quote. Ransomware attacks soared last year, pushing demand for coverage after the pandemic-induced work-from-home era, also made remote workers more vulnerable to digital attacks. Those attacks also spurred companies and individuals to adopt more robust cybersecurity measures. Loss ratios for standalone and package policies settled down to 43 and 48 percent, respectively. Am Best said in signs of life for the industry after incurred losses had increased in 2020 and 2021. Surplus lines insurers which patch unique risks

uncovered by conventional policies are also gaining popularity, end quote. Spotify looks to have completely shifted its podcasting strategy. Instead of trying to be the HBO or Netflix for audio with lots of exclusive programming behind the paywall, they seem to be pivoting to providing tools for podcast creators. So basically, if you want to use a proper analogy, they seem to be turning into the YouTube for podcasters, going Bloomberg. On July 5th, the company announced that it was laying off 2% of its workers or 200 people following a previous round of cuts in January. and that the job cuts would primarily be in its talk division. Zahar Elhabashi, vice president for the podcast business, called the latest losses a, quote, difficult but necessary decision that will allow the company to enter its next chapter in podcasting. That chapter will put a stronger emphasis on providing tools to people who produce podcasts and then on making money from advertising with a model that has echoes of YouTube. This is a departure from much of what Spotify had been building,

which was a sort of HBO of podcasts where it used premium content, to persuade people to use its platform. Spotify has even begun to distribute its high-end podcast by Emma Chamberlain, Anything Goes, on other platforms such as Apple Podcasts. It's in conversations to do the same with Dax Shepard's armchair expert. As part of this shift, two of Spotify's early and prominent studio purchases, Gimlet Media and Parkast, are losing their respective branding and joining the broader umbrella of Spotify Studios. The podcast studios hadn't been allocated their annual budgets so they couldn't approve new shows. Deals weren't renewed, and many top stars such as Brock and Michelle Obama, Bray Brown, and Esther Perel, allowed their deals to lapse.

Almost all the founders of the acquired companies had also left Spotify, amid the widespread uncertainty rumors circulated among employees about a looming reorganization. While it's retrenching on some of its big investment, Spotify is pegging its podcasting future more firmly on another acquisition it made during its free spending stage. Megaphone, the podcasting, hosting, and advertising service, that it bought for $235 million in November 2020. That comes with its own set of expenses in late June. Spotify is flying podcasters and musicians out to the French Riviera for a party at Con Lyon, a crucial event for any company hoping to get some FaceTime with the world's biggest ad agencies, end quote. Another day, another one of these, meta has detailed IJEPA,

a computer vision model that uses common sense world knowledge to create more accurate images, avoiding errors like, you know, hands with extra digits. Quoting Silicon Angle, artificial intelligence researchers from meta platforms say they're making progress on the vision of its chief AI scientist, Jan Lecun, to develop a new architecture for machines that can learn internal models of how the world works. The idea is that such an architecture would help AI models learn faster,

plan how to accomplish complex tasks, and readily adapt to unfamiliar situations. Meta's AI team said, Today, it's introducing the first AI model based on a component of that vision, called the Image Joint Embedding Predictive Architecture, or IJEPA. It's able to learn by creating an internal model of the outside world that compares abstract representations of images as opposed to comparing the pixels themselves. That means it learns in a way that's much more similar to how humans learn new concepts. IJEPA is based on the idea that humans learn massive amounts of

background information about the world as they passively observe it. It basically attempts to copy this way of learning by capturing common sense background knowledge of the world and encoding it into digital representations that can be accessed later. The challenge is that such a system must learn these representations in a self-supervised way, using unlabeled data such as images and sounds as opposed to labeled datasets. At a high level, IEPA can predict the representation of part of an input, such as an image or piece of text, using the representation of other parts of that same input. That's different from newer generative AI models, which learn by removing or distorting portions of the input, for instance, by erasing part of an image or hiding some words

in a passage, then attempting to predict the missing input. According to Meta, one of the shortcomings of the method employed by generative AI models is that they try to fill in every bit of missing information, even though the world is inherently unpredictable. As a result, generative methods often make mistakes a person would never make because they focus too much on irrelevant details. For instance, generative AI models often fail to generate an accurate human hand, adding extra digits or making other errors. IJEPA avoids such mistakes by predicting missing information in a more human-like way, making use of abstract prediction targets in which unnecessary pixel-level details are eliminated. In this way, IJEPA's predictor can model spatial

uncertainty in a static image based on the partially observable context, helping it predict higher-level information about unseen regions in an image, as opposed to pixel-level details. Meta-Setat IJEPA has displayed a very strong performance on multiple computer vision benchmarks showing itself to be much more computationally efficient than other kinds of computer vision models. The representations it learns can also be used for other applications without needing extensive fine-tuning, end quote. Finally today, Comprehensive.io says salaries for senior engineers who specialize in artificial intelligence and machine learning are 12% higher than for those who do not. AI-related salaries altogether rose 4% in just the

past year on top of that. According Bloomberg, right now, the average salary for a senior software engineer who specializes in artificial intelligence or machine learning is 12% higher than the salary of one who doesn't, says Roger Lee, who runs both layoff tracker layoffs.FY and Comprehensive.io, which monitors salaries in the industry. Although pay for tech workers has plateaued or even dipped, in the past year, salaries of AI-related roles have risen 4%, Lee says. In the startup world where venture funding is largely drying up and many start are cutting jobs to survive, investors are still salivating over companies offering anything AI-related.

This is especially true of generative AI, the term encompassing technology such as OpenAI's ChatGBT. At larger companies, workers who attach themselves to AI projects may be the safest, at least in the short term. As Drew Houston, Dropbox's chief executive officer announced that the company was cutting 500 workers, he noted that part of the rationale for the cuts was to make way for AI-related hires. Our next stage of growth requires a different mix of skill sets, particularly in AI and early stage product development, end quote. When Mark Zuckerberg laid out plans in March to eliminate 10,000 jobs at meta platforms, he said one company focus was, quote, building AI tools to help engineers write better code faster,

enabling us to automate workloads over time, end quote. At the same time that AI promises to be Silicon Valley's next big boon, it also threatens to upend the conventional wisdom that simply learning computer programming is insurance against job insecurity. IBM CEO Avrend Krishna recently told the Bloomberg news that he could, quote, easily see thousands of IBM jobs being replaced by AI in the next five years, particularly for mundane tasks in human resources. He later said he thought that AI would end up creating more jobs than it destroyed. Right now, AI may actually be serving as a face-saving explanation for executives who want to avoid admitting that job cuts are a result of overhiring during the pandemic boom.

There's no evidence that AI has already begun to displace significant numbers of technical jobs. Still, it's easy for an engineer to watch ChatGPT spit out code and correct its own programming errors. And imagine how a team of five programmers could be replaced by two humans and some advanced AI tools, end quote. This is actually from a larger piece about the larger job market in tech at the moment, so maybe worth clicking through to read the whole thing. Nothing for you today, unless you want to hear more about Zelda, where I found a way to paste two fans together with a steering stick that makes flying around, especially in the depths super easy. You just put a bright bloom seed on the front of it. So now I'm going around trying to get as much zonite as I can to max out my battery. If that's gobbly gook to you,

you know, this is how I'm spending my time these days. Talk to you tomorrow.