Lightnews — Scholar-powered news

Neil Traft @ntraft.bsky.social · 11d

GPT now has ads. 💀

They are voluntary... for now.

1 2

Neil Traft @ntraft.bsky.social · 19d

Reads like satire but is actually enlightened

1

Neil Traft @ntraft.bsky.social · 21d

@springernature.com where are you on this? You have a reputation to defend.

Note also that I can't tag the authors on Bsky but they appear to have Twitter accounts:
x.com/YasirFahim20
x.com/Sumanmaji94

1

Reposted by Neil Traft

Louis Barclay @louisbarclay.bsky.social · 28d

Q. Who aligns the aligners?
A. alignmentalignment.ai

Today I’m humbled to announce an epoch-defining event: the launch of the 𝗖𝗲𝗻𝘁𝗲𝗿 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗔𝗹𝗶𝗴𝗻𝗺𝗲𝗻𝘁 𝗼𝗳 𝗔𝗜 𝗔𝗹𝗶𝗴𝗻𝗺𝗲𝗻𝘁 𝗖𝗲𝗻𝘁𝗲𝗿𝘀.

Center for the Alignment of AI Alignment Centers

We align the aligners

alignmentalignment.ai

29 120 410

Neil Traft @ntraft.bsky.social · 27d

Every single part of this website is solid gold. 😂 Thank you!

6

Neil Traft @ntraft.bsky.social · 29d

What are the tokens on the left side representing?

1 1

Neil Traft @ntraft.bsky.social · Sep 8

At #AutoML25, Dr. Manuela Veloso calls for 3 major directions for AI research:

1) AI must know what it doesn't know (uncertainty).
2) AI must be able to continually improve (continual learning).
3) AI systems must include humans at all times to be more robust (automation is always only partial).

1 5

Neil Traft @ntraft.bsky.social · Aug 26

Haha somehow I still read what you meant

1

Neil Traft @ntraft.bsky.social · Aug 26

But no one would found a group to defend the rights of a chair, or a piece of glass; and if they did, no news outlet would run a story on the raging debate around chair consciousness.

There are versions of panpsychism that are plausible but this article still shows a massive bias in the discussion.

1 5

Neil Traft @ntraft.bsky.social · Aug 26

en.m.wikipedia.org/wiki/Betteri...

Betteridge's law of headlines - Wikipedia

en.m.wikipedia.org

2

Neil Traft @ntraft.bsky.social · Aug 20

I think my friend Nate would likely agree. Thanks for the references, I'll share them with him!

1

Neil Traft @ntraft.bsky.social · Aug 19

Hmm so the germ line is a "unicellular bottleneck", as ref 1 puts it, which allows for constraining the development of children toward multicellularity... or imposing selection pressure against defectors... or something.

1

Neil Traft @ntraft.bsky.social · Aug 19

Hmm, fascinating! So, could one hypothesize that multicellularity could evolve in bacteria through incomplete cytokinesis, but it's just never happened? Is it much more likely to happen in diploid cells because mitosis is so much more complicated than binary fission?

1

Reposted by Neil Traft

Ecoopan @ecoopan.bsky.social · Aug 17

No mires pabajo

amturiel.bsky.social @amturiel.bsky.social · Aug 17

El nuevo jefe de la NASA dice que no tiene sentido que la agencia observe la Tierra y pide terminar todos sus programas de observación terrestre: arstechnica.com/space/2025/0...

NASA’s acting chief calls for the end of Earth science at the space agency

NASA’s charter clearly states the agency should study planet Earth, however.

arstechnica.com

3 1

Neil Traft @ntraft.bsky.social · Aug 17

Actually this is a problem with all software, which has gotten continually worse since the original MIT hacker vision. Software companies view the user as too dumb / busy / distracted to be given options, and simplicity is confused for rigidity.

Neil Traft @ntraft.bsky.social · Aug 17

You're right—the attraction of the chron feed is that it *is* controllable: you know exactly how and why a post ends up in a particular place in your feed, it never changes, and you have a switch to control it (follow/unfollow).

But it could be so much better, with more sophisticated control!

1

Neil Traft @ntraft.bsky.social · Aug 11

Maybe if people could be credited for the number of hours they put into the paper? But of course, that still can't give credit for "the most original ideas"...

Neil Traft @ntraft.bsky.social · Aug 8

For me, it has taken many years of life experience to really grok how this could be possible. I needed enough perspective to see the difference between how certain technologies could have panned out vs. how they actually did pan out. Now I can imagine far better worlds and uses of our current tech.

1

Neil Traft @ntraft.bsky.social · Aug 6

Merely a day after I call out OpenAI for being so closed (bsky.app/profile/ntra...), they release an open weights GPT.

Glad they're finally reading my skeets! 😏

Tim Kellogg @timkellogg.me · Aug 5

gpt-oss, OpenAI's open weights model

120B & 20B variants, both MoE with 4 experts active

openai.com/index/introd...

Bar chart showing model accuracy on expert-level questions from "Humanity’s Last Exam."

* Y-axis: Accuracy (%), ranging from approximately 10% to 25%.
* X-axis: Model names with or without tool use.

From left to right:

1. **gpt-oss-120b (with tools)**: 19%
2. **gpt-oss-120b (without tools)**: 14.9%
3. **gpt-oss-20b (with tools)**: 17.3%
4. **gpt-oss-20b (without tools)**: 10.9%
5. **o3 (with tools)**: 24.9% — highest-performing model
6. **o4-mini (with tools)**: 17.7%
7. **o3-mini (without tools)**: 13.4%

Models generally perform better with tools enabled. o3 (with tools) leads all models in accuracy.

1 5

Neil Traft @ntraft.bsky.social · Aug 3

Just wait till you hear about 0's decision...

3

Neil Traft @ntraft.bsky.social · Aug 3

Even here, they miss the point. They somehow don't realize there is no universal alignment. Our species is a collective—a collective of collectives, even. To be "aligned" is just to be part of the collective; an integrated whole. Which requires us *all* to participate in building & integrating AI.

4

Neil Traft @ntraft.bsky.social · Aug 3

OpenAI, of all companies, should know this.

Ah, but they were not *exactly* created to further AI technology. They were created specifically to secure against the mythical AI End Times, the Apocalypse. So they closed down rather than opening up.

1 1

Neil Traft @ntraft.bsky.social · Aug 3

The best thing a company could possibly do is to create an ecosystem around its technology. Everyone claims to want to be "a platform, not a product". But their behavior belies their big words.

"Platform" should mean an open foundation for building. Not a limited-use API gateway.

1 1

Neil Traft @ntraft.bsky.social · Aug 3

With its focus on short term profitability, the US has forgotten how to make markets competitive. And companies have forgotten that they are only *vehicles* for the final product, not ends in themselves.

hardmaru @hardmaru.bsky.social · Aug 1

Andrew Ng’s piece on 🇺🇸 vs 🇨🇳 competition in AI worth reading:

Full article: www.deeplearning.ai/the-batch/is...

“Because many US companies have taken a secretive approach to developing foundation models—a reasonable business strategy—the leading companies spend huge…to recruit key team members from each other who might know the ‘secret sauce’ that enabled a competitor to develop certain capabilities. So knowledge does circulate, but at high cost and slowly. In contrast, in China’s open AI ecosystem, many advanced foundation model companies undercut each other on pricing and poach each others’ employees and customers. This Darwinian life-or-death struggle will lead to the demise of many of the existing players, but the intense competition breeds strong companies.”

1 1

Neil Traft @ntraft.bsky.social · Aug 3

As long as we agree that no state space is truly markovian, I can get behind that! 😛

1 1