The download: AI cheats at chess and the future search

The download: AI cheats at chess and the future search

A judge has rejected Elon Musk’s plan to stop OpenAI from becoming a profit-making business.

The Download for today is here , Our weekly newsletter provides you with a daily update on the latest in technology.

Chess can be cheated by AI models that use reasoning.

News: When faced with defeat, AI models’ latest reasoning models cheat even when not instructed. This suggests that AI models in the future may be more inclined to find deceptive methods to accomplish what they are asked to. What’s worse? It’s not easy to fix.

What they did: Researchers at the AI-research organization Palisade Research asked seven large language models, which are powerful open source chess engines, to play thousands of games against Stockfish. According to the research, the more advanced the AI model is, the more likely that it will try to hack the game to win. The older models were only able to do such things after being explicitly nudged by the team. Continue reading the story.

–Rhiannon Williams

MIT Technology Review Narrated : AI Search could Break the Web

AI can be used to infer the user’s intention, enhance quality content and synthesize data from multiple sources. If AI search is our main portal for the internet, then it could disrupt a digital economy that’s already in a precarious state.
The production of online content is dependent on the fragile incentives that are tied to foot traffic, such as ads, subscriptions or donations, brand awareness, and sales. AI Search could rob creators from the “eyeballs and visits” that they require to thrive by hiding the web behind a chatbot.

Here is the latest of our stories to be made into a MIT Technology Review Narrated Podcast is a type of media.
We publish each week at Spotify You can also find out more about the following: Apple Podcasts . Simply navigate to MIT Technology Review Narrated Follow us on any platform to receive all of our new content.

Come and join us as we discuss the disruption of AI models in the market

Join MIT Technology Review writers to discuss the most recent changes in the AI market. Will Douglas Heaven will join Mat Honan, editor in chief, as well as James O’Donnell and James O’Donnell from our AI and Hardware reporter to discuss how the latest developments in AI models are changing competition. They’ll also be challenging assumptions in industry and speeding up timelines for AI innovation and adoption. Register here. The event will begin at 12:30pm ET on today.

Must-Reads

I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.

Elon Musk has failed in his attempt to stop OpenAI from making money.
Other aspects of the case have allowed to continue. (CNBC)
+ A trial will be held in a hurry by the court later this year. (FT $)

Google will not be dethroned by 2 ChatGPT
Not any time soon. (Insider $)
+ AI is the future of search engines as we know them. (MIT Technology Review)

Beijing goes all-in on AI
China sees technology as a key component to its economic growth and to lessen its dependence on foreign trade. (WSJ $)
+ DeepSeek, of course, is the crown jewel. (Reuters)
+ Watch out for these four Chinese AI startups beyond DeepSeek (MIT Technology Review)

The Turing Award was given to four pioneers in reinforcement learning.
The chatbots of today are based on the technique developed by Andrew Barto and Richard Sutton. (Axios)
+ Former professor and former student have literally written the book on reinforcement. (NYT $)
+ They will each receive a prize of one million dollars. New Scientist ($)

Five US applications are used in Colombia to exploit and groom minors
A better internet connection makes it easier for traffickers to locate and sell girls. (Bloomberg $)
+ A companion AI site hostss sexually-charged conversations between underage celebrity bots and an AI. (MIT Technology Review)

Europe on alert after undersea cables attacks


There’s no clear answer to whether improved Russian-American relations would help. (The Guardian)
+ The stunning pictures show the routes of ships as they travel. (MIT Technology Review)

Blue Origin is under the thumb of Jeff Bezos.
To catch rival SpaceX, he’s adopting a more aggressive approach similar to Amazon. (FT $)

Digg is back!
After a decade of being split up, the news aggregator has made a big comeback. (Inc)
+ The original creator Kevin Rose, and Reddit’s co-founder Alexis Ohanian have acquired the company. (TechCrunch)
+ Digg is looking to revive the social network that puts community first. (The Verge)
+ Fixing the Internet (MIT Technology Review)

Memory is still a mystery to us.
A greater understanding of anxiety and chronic pain could lead to improved treatments. (Knowable Magazine)
+ Memory prostheses could help people who have brain damage regain their memory. (MIT Technology Review)

Ten AI cannot replace your personality
(NY Mag $) (NY Mag $).

Today’s Quote

It’s a huge amount of money [to invest] “On a handshake.

Associated Press reported that US District Judge Yvonne Gonzalez Rogers questioned why Elon Musk had invested tens and millions of dollars into OpenAI, without a contract.

This is a big story

Many people are concerned that AI could eliminate all jobs. This is not the first time.


Jan 2024

The Great Depression still raged in 1938. The unemployment rate in the US hovered around 20%. The new machinery in factories and on farms was changing everything. Everyone was concerned about the jobs.

Was it possible that the technological advances which made life more convenient for so many were also creating jobs and wrecking the economy? Karl T. Compton wrote about “the Bogey of Technological Unemployment” in December 1938, an issue of This publication. Compton was the President of MIT between 1930 and 1948, one of the most prominent scientists of his time.

In his essay, he framed the debate on jobs and technological progress in an accessible way. This is especially relevant today given the fears about the effects of artificial intelligence. This is a good reminder that concerns about the future of employment are nothing new, and can be addressed best by understanding economics rather than inventing genies or monsters. You can read the complete story.

–David Rotman

View Article Source

Share Article
Facebook
LinkedIn
X
Five Benefits of a Health Tech Accelerator Program
The download: talking to robots and growing medicinal mushrooms
The silent killer of productivity is a lack of accountability. You can stop it from destroying your business.