Matthew Sheffield

🟦 Host, @TheoryChange
🟦 As seen in: @discoverflux, New York Times, Washington Post, NBC, CNN, etc.
🟦 Cognitive science, AI, philosophy, media, politics...

[Matthew Sheffield @mattsheffield]

Thousands of people think their favorite chatbot has a soul, but why does no one ever say the same about Midjourney?

My latest AI and philosophy column is out now. Link in next.

[Matthew Sheffield @mattsheffield]

Here it is: "Large language models and the textual nude"

https://t.co/a0EI428O4o

[Matthew Sheffield @mattsheffield]

Thousands of people think their favorite chatbot has a soul, but why does no one ever say the same about Midjourney?

My latest AI and philosophy column is out now. Link in next.

[AI at Meta @AIatMeta]

We’re sharing the next major milestone in our non-invasive brain-to-text decoder research: Brain2Qwerty v2.

Building on v1, which was published today in @Nature, Brain2Qwerty v2 is the highest-performing end-to-end pipeline capable of real-time sentence decoding from raw brain signals. It advances beyond character-level performance to decoding words and semantics, enabling accuracy for overall communication.

We believe this research has the potential to make a real difference for the millions of people who suffer from brain lesions or disorders that prevent them from communicating.

🧵👇

[Alec Stapp @AlecStapp]

Very, very important caveat:

[60 Minutes @60Minutes]

“We spotted nine Polymarket accounts, all connected, who made, collectively,$2.4 million betting almost exclusively on U.S. military operations,” says Nicolas Vaiman, co-founder of the small data analytics firm Bubblemaps.

“And now here's the crazy part: 98% win rate.” https://t.co/wrTjsWhXxW

[Luiza Jarovsky, PhD @LuizaJarovsky]

🚨 Google published its AI policy recommendations to the U.S. government, and they're based on a rigid separation between "frontier AI" and "widely-deployed AI applications"

It states, for example:

"The issues raised by the widespread use of AI applications like chatbots are distinct from the kinds of national security issues posed by advanced frontier AI models."

The main problem is that these two categories are NOT rigidly separated.

The most popular and widely used chatbots today are ChatGPT, Gemini, Claude, and Copilot, and they are ALL based on frontier AI models.

Yes, the vast majority of people are not using AI chatbots for high-stakes missions or posing a national security risk.

But the models with the highest penetration are the same ones that could enable a major attack or incident.

This separation is not functional and should not be used by any government to determine the applicable AI law or policy framework, as it will not deter national security threats.

At the model level, regulation should be flat and cover the highest possible risk.

At the application level, there should be more nuance, but these should affect usage and privacy policies, guardrails, and controls.

👉 I've recently written about the current state of American AI policy in light of recent developments. Read my full article below.

👉 To learn more, you can also join my upcoming training program on global AI policy, ethics, and regulation.

[Scott Lincicome @scottlincicome]

New research from @nytimes finds that the US government is actively doing "critical minerals" deals with FOURTEEN different companies that have financial ties to the Trump and/or Lutnick families - deals worth around $9 billion in all:

[Matthew Sheffield @mattsheffield]

This study is pretty damning also for people like Chomsky and Pinker, who have a notion of a "language module" that exists somewhere invisible within the brain. But if a non-specialized program and brains both use relationality, then there is no evolutionary magic module.

[Matthew Sheffield @mattsheffield]

This is a fascinating study arguing that language concepts in brains are not translated from some supposed inner language.

Rather, they seem to be spatially stored and relationally used. There is no "language of thought" only processual tokens. https://t.co/i3NrTmhxlw

[Matthew Sheffield @mattsheffield]

Meaning is enacted rather than discrete. No particular cell has "the" copy of what something means. Info is stored in many places, unknown how many, and thoughts are assembled based on relationalities.

This even holds for animals like planarians which can regenerate brains.

[Greg Olear @gregolear]

The cat's out of the bag, guys: my new Epstein book, ANOTHER WONDERFUL SECRET, is now available.
https://t.co/uSI4Er2OZB

I did a little Q&A about the Q&A book, in which I answer my own burning questions:
https://t.co/uge49NDzu6

[Rob Henderson @robkhenderson]

“The only trait that consistently predicted objections to remote work was narcissism…The higher the opinions of themselves leaders expressed, the more they coveted power and status — and the more they favored return-to-office mandates.” https://t.co/cmiPR6QLd8

[Hemant Mehta @hemantmehta]

Why is he threatening us with a good time?

[Matthew Sheffield @mattsheffield]

@cit_collins That's a good way of putting it. No one can have all positive attributes, including the person doing the judging. There is no ideal person out there, only people that one can mesh with and support/be supported.

[Matthew Sheffield @mattsheffield]

Online dating has been often observed to decrease desire to enter a formalized relationships, but a study just published indicates that large amounts of dating experience also leads to long-term commitment failure, as some people develop unrealistic expectations of partners.

[Matthew Sheffield @mattsheffield]

@HeiderKate Enterprise customers already could get the updates, so this is just applying it to regular consumers.

As stated in the piece, the motivation seems to have been the fact that unpatched Win10 machines are virus magnets.

[Matthew Sheffield @mattsheffield]

As with Windows XP, Microsoft is having to bend to the reality that people just want to keep using Windows 10. Now the company is giving another year of free updates https://t.co/1d0p68gBpp

[Matthew Sheffield @mattsheffield]

Source: https://t.co/3EKF6xCS4m

[Matthew Sheffield @mattsheffield]

Source: https://t.co/59PF97M09z

[Matthew Sheffield @mattsheffield]

Since launching his Iran war, Trump's standing has fallen in every country worldwide according to a new Pew Research global survey.

Compared to Biden, Trump's ratings are lower in every country except Israel and Hungary.