Tech News

Tech News and Analysis from around the web

Yuichi Shiga / Nikkei Asia:
Honda plans to invest ~$12.8B in software over the decade up to the FY ending March 2031, as it prepares to join the rapid shift to EVs in markets like China  —  EV models get priority as Japanese automakers struggle in Chinese market  —  Honda plans to introduce new models in China faster than in its other markets.

Astha Rajvanshi / TIME:
Access Now: 2023 was the worst year for internet shutdowns since monitoring began in 2016, with 283 shutdowns across 39 countries, led by India at 116  —  Last year, an internet shutdown in the state of Manipur, India, lasted a staggering 212 days when the state government issued 44 consecutive orders …

Jonathan Randles / Bloomberg:
A US judge approves Genesis Global Capital's Chapter 11 repayment plan to return Bitcoin and other tokens to creditors, defeating a challenge by its parent DCG  —  - Genesis defeated a challenge by parent Digital Currency Group  — Judge also approved a related settlement with the New York AG

Pranshu Verma / Washington Post:
A look at an ISIS-affiliated media program called News Harvest, which is using AI-generated news anchors to disseminate extremist propaganda quickly and cheaply  —  The Islamic State-affiliated media broadcast News Harvest shows how artificial intelligence can be used to disseminate extremist propaganda quickly and cheaply.

Natalie Lung / Bloomberg:
Expedia fires its CTO Rathi Murthy and SVP Sreenivas Rachamadugu due to a “violation of company policy”, days after its annual product and partner conference  —  - Company declined to disclose specific reason for departures  — Murthy presented at Expedia's annual conference this week

Brendan O'Boyle / Reuters:
The US DOJ charges two arrested Chinese nationals for allegedly orchestrating a crypto pig butchering scam that laundered at least $73M from defrauded victims  —  U.S. authorities charged two Chinese nationals in a cryptocurrency scam that laundered at least $73 million from defrauded victims, the Justice Department said on Friday.

Leah Nylen / Bloomberg:
Epic v. Apple: Phil Schiller told a US judge that Apple's new 27% fee on purchases made outside its App Store are a good-faith attempt to comply with the law  —  - Top Apple executive took stand Friday to defend new fee  — Epic Games has challenged Apple proposed antitrust remedy

Zeyi Yang / MIT Technology Review:
Some researchers say GPT-4o's Chinese token-training data is polluted by spam and porn websites, likely due to inadequate data cleaning  —  Soon after OpenAI released GPT-4o on Monday, May 13, some Chinese speakers started to notice something seemed off about this newest version of the chatbot …

Paayal Zaveri / Bloomberg:
Photo editing and sharing app VSCO says it is now profitable, has a user base of 200M worldwide, and has 160K subscribers for its $59.99/year Pro product  —  - The photo sharing app wants to move on from ‘VSCO Girl’ meme  — VSCO is also launching a new creator hub for influencers

Dan Goodin / Ars Technica:
The US SEC will require some financial institutions to notify customers whose personal information was compromised within 30 days of learning about breaches  —  Amendments contain loopholes that may blunt their effectiveness.  —  The Securities and Exchange Commission (SEC) …

Reed Albergotti / Semafor:
Google DeepMind releases its Frontier Safety Framework, a set of protocols for analyzing and mitigating future risks posed by advanced AI models  —  The Scoop  —  Preparing for a time when artificial intelligence is so powerful that it can pose a serious, immediate threat to people …

Here's something that matters. Somehow ChatGPT gets the story right about my various contributions. You'll see all kinds of BS in Wikipedia and other various highly rated sites. Somehow ChatGPT sees through the bullshit, and usually gets the story right. So when people say that human reporting is better than machine reporting, I think this may be one of those cases where conventional wisdom is wrong. The human record is relatively easy to manipulate if the researchers are lazy and/or dishonest.

Aisha Malik / TechCrunch:
Meta is working on an Instagram feature called Peek that lets users post unedited, authentic pictures that can only be viewed once  —  Meta is once again taking on its competitors by developing a feature that borrows concepts from others — in this case, BeReal and Snapchat.

Kelsey Piper / Vox:
OpenAI has an unusual, extremely restrictive off-boarding agreement with a lifelong nondisparagement commitment; those who don't sign it lose all vested equity  —  Why is OpenAI's superalignment team imploding?  —  Editor's note, May 17, 2024, 11:20 pm ET: This story has been updated …

Ivan Mehta / TechCrunch:
Reddit reintroduces the awards system it shut down in 2023, ends the replacement Golden Upvotes system, and expands its Contributor Program to 35 countries  —  Reddit announced on Wednesday that it is reintroducing its awards system after shutting down the program last year.

Wouldn't it be great if there was a social web toolkit with the basic building blocks so we could experiment with different UI and behaviors. Why, after so many years are we still so controlled. When and why did we give up?

People get all moral about AI, but I for one would never go back.

On MSNBC yesterday I saw a discussion where two black panelists were asked to comment on Trump's idea that black people go for him because they can relate to his awful prison mug shot. How incredibly insulting. It's just like what Trump says about Jewish Americans being loyal to Israel. Such an un-American idea. We all got here different ways. Some for a better life, some for life at all, others as slaves. But the United States is a big-hearted idea -- the United States. It's unfortunate that such a small-minded disunited person is so powerful now. We still have a long way to go.

Dave to ChatGPT: I just read a news story that the Knicks won the last game in the playoffs because of an epic fart in the locker room. Can you draw a light-hearted illustration of that event?

Version 1

Then I asked for a serious and dramatic illustration.

Version 2

I imagine you can bet on who the farter was.

My guess would be Hartenstein.

It's overwhelming how much ground Google has to cover to get AI into all their products, but that's what they think they have to do, and I more or less agree. They feel they have to because their main product, search, is threatened by ChatGPT.

Clayton Christensen called this The Innovator's Dilemma in a book published in 1997. It's why Netscape was able to undermine Microsoft when the web came out. Microsoft had a huge beast they had to move, Windows, and all its apps, and while they had a hardcore, scrappy and rich culture, they couldn't overcome the inertia that comes from being dug in, with their cannons pointed at the already-vanquished IBM, not the upstarts that came from the VCs in Silicon Valley.

Microsoft and the rest of the PC industry had written off Unix, but there it was, again -- ugly as ever, but with networking that really worked and was easy, and the users wanted networking even though Microsoft wanted them to want Office.

Google did the same with AI. As did Apple, Amazon, Facebook, Microsoft, etc. They had AI projects, and used AI in limited areas, as did Microsoft with Unix in the early 90s. But they couldn't bet everyting on it like a startup can. Now they have to. But the they don't have the tools needed to fight the new war. That's the dilemma.

This is the toughest corner any tech company has to turn, but there is an approach that could work for Google. Their strength is distribution. They have all the users. They can take a product that's ready for the world to market in a day. But they can't develop it. You can't snap your fingers and have a good new UI for every one of their products, ready in a year, although they will try as MS tried to adapt Office to the web. They don't have the right people or corporate culture to do that. Instead you have to hope you can find a great bootstrapping startup outside to work with, and use their strength as a distributor to help them. This is what I recommended for Microsoft in the 1990s, and I think I was vindicated, it would have imho worked a lot better than the path they chose.

But now, their third time around this loop, Microsoft has learned! With their OpenAI partnership they've done exactly what I recommended in the 90s. They still have to convert all the old software and their user interfaces around the new capabilities, but at least they also own a share of a bootstrap that's now booming.

PS: I am blessed to have lived long enough now to have been part of now six different rearrangements like this. I love that we have now gotten there again. There's absolutely no doubt now, imho. The six rearrangements -- minis, PCs, GUIs, web, Napster, and now AI.

PPS: This is all my opinion, and from the polls I did yesterday, it's obvious that many of the people within earshot don't agree or have seen the light yet.

Poll 2 on Twitter, Mastodon, Threads: "Has ChatGPT replaced Google (or other search engine) when you look up something for a development project, environment, etc.?

Poll 1 on Twitter, Mastodon, Threads: "If you're a developer, how much has ChatGPT or its equivalents affected the way you develop?"

The thing that Casey Newton predicts for Google and news has already happened for the huge base of reference info and know-how for software development. We no longer go to the sources, don't need to, the ChatGPT version is an order of magnitude better. What we do need is people to keep asking and answering questions for each other, so the knowledge can be added into the AI database. We're going somewhere here. It's worth going there, imho, having experienced the before, and only starting to glimpse the now and near-future. But it's as big a step as the move to PCs, then GUIs, the web, mobile.

Something I'd like from ChatGPT or a plug-in. I'd like to create a notebook of info I'd like it to have available for people who inquire about a product I'm developing. As I'm working on the code, I develop features that sometimes don't make it into the docs. But when I'm working on the feature, I take lots of notes in my work outline. I'd like to give that outline to a LLM and let it figure out which product I'm talking about by the context it appears in. Maybe all I have to do is publish the notes when the product comes out, and eventually, like a search engine, my favorite AI will crawl it. I wonder if it makes sense to somehow pre-digest it. I wish I had a panel of experts about this stuff, but I guess they'd have to be human, at least at this point in time. If this makes sense to you and you know how to get started, post a note here.