Author HubMatthias Bastian

Matthias is the co-founder and publisher of THE DECODER, exploring how AI is fundamentally changing the relationship between humans and computers.
AI in practice

Cloudflare CEO Matthew Prince sees trouble ahead for the open web

AI research

Wait a minute! Researchers say AI's "chains of thought" are not signs of human-like reasoning

AI in practice
Update

US Copyright Office says fair use does not cover AI trained on "vast troves of copyrighted works

Ad
AI in practice
Short

OpenAI's recent move to tweak its planned for-profit restructuring hasn't changed Elon Musk's stance. According to Musk's attorney Marc Toberoff, the adjustment is just a "transparent dodge" and does nothing to address concerns that OpenAI's nonprofit assets are still being used to benefit private interests—including Sam Altman, investors, and Microsoft. Musk previously tried to block the restructuring in court and made a $97.4 billion bid to take over OpenAI, but those efforts failed. Parts of the lawsuit are still ongoing.

The founding mission remains betrayed.

Marc Toberoff, Musk’s lead counsel

Ad
Ad
AI in practice
Short

Anthropic has introduced a new "AI for Science" initiative that offers up to $20,000 per month in API usage credits to selected researchers. According to the company, applicants are evaluated using objective criteria and must pass a biosecurity review. The program is open to individuals over 18 and research teams from academic institutions, except applicants from countries including China, Russia, and Iran. Anthropic says it may also reject applications that violate its rules. Full details and eligibility requirements are available on the Anthropic website.

AI in practice
Short

New York City's Metropolitan Transportation Authority (MTA) plans to use AI to automatically detect suspicious activity on subway platforms and alert police. According to MTA head of security Michael Kemper, the goal is "predictive prevention." The software analyzes live feeds from surveillance cameras, but will not use facial recognition, MTA spokesperson Aaron Donovan said. The move comes after a series of attacks on the subway. Civil liberties groups, including the NYCLU, have criticized the plan as excessive. NYCLU policy counsel Justin Harrison warned that AI systems are prone to mistakes and could worsen existing inequalities. The MTA has now installed surveillance cameras on every subway platform and inside every train car, with about 40 percent of them monitored in real time.

Ad
Ad
AI in practice
Short

How useful are million-token context windows, really? In a recent interview, Nikolay Savinov from Deepmind explained that when a model is fed many tokens, it has to distribute its attention across all of them. This means focusing more on one part of the context automatically leads to less attention for the rest. To get the best results, Savinov recommends including only the content that is truly relevant to the task.

I'm just talking about-- the current reality is like, if you want to make good use of it right now, then, well, let's be realistic.

Nikolay Savinov

Recent research supports this approach. In practice, this could mean cutting out unnecessary pages from a PDF before sending it to an AI model, even if the system can technically process the entire document at once.

AI in practice
Short

Does saying "please" and "thank you" really help when talking to AI? According to Murray Shanahan, a senior researcher at Google Deepmind, being polite with language models can actually lead to better results. Shanahan says that clear, friendly phrasing—and using words like "please" and "thank you"—can improve the quality of a model's responses, though the effect depends on the specific model and the context.

There's a good scientific reason why that [being polite] might get better performance out of it, though it depends – models are changing all the time. Because if it's role-playing, say, a very smart intern, then it might be a bit more stroppy if not treated politely. It's mimicking what humans would do in that scenario.

Murray Shanahan

Ad
Ad
AI in practice
Short

Anthropic employees are about to get rich. The company is giving current and former staffers who have been with the company for at least two years a chance to cash out up to 20 percent of their shares—capped at $2 million per person. Anthropic will buy back the shares at its latest $61.5 billion valuation, matching the price from its March funding round. The buyback, worth several hundred million dollars in total, is expected to wrap up by the end of the month, according to The Information. Anthropic, founded by former OpenAI researchers, now has more than 800 employees.

Google News

Bank details

IBAN: DE87 1203 0000 1086 0070 75
Account holder: DEEP CONTENT GbR
Purpose: Support THE DECODER
AI in practice

OpenAI launches new ChatGPT agent that automates complex tasks for Pro, Plus, and Team

AI in practice
Update

Kimi-K2 is the next open-weight AI milestone from China after Deepseek

AI research

New Energy-Based Transformer architecture aims to bring better "System 2 thinking" to AI models

Google News