Mark Needham (@mbneedham) header
Mark Needham (@mbneedham) profile picture
Joined September 2011
Photo by Mark Needham (@mbneedham)Photo by Mark Needham (@mbneedham)Photo by Mark Needham (@mbneedham)Photo by Mark Needham (@mbneedham)Photo by Mark Needham (@mbneedham)Photo by Mark Needham (@mbneedham)Photo by Mark Needham (@mbneedham)Photo by Mark Needham (@mbneedham)
To see is to forget the name of the thing one sees. — attrib. Paul Valéry
3
1
28
Mark Needham retweeted
Look at the signatories! xitter.cf/satyanadella/status/20…
Open-weight models are essential to a healthy AI ecosystem. Together with others across our industry, we are outlining a path for open-weight models to strengthen American competitiveness and expand economic opportunity, while protecting national security. microsoft.com/en-us/corporat…
1
2
34
Marc Cohodes on Nvidia's Jensen Huang $NVDA youtube.com/watch?v=FDJTUJm1…
28
88
20
435
157,593
Mark Needham retweeted
Frustrating how many on the left are falling for this. "Open" models are still weapons of capital, created by plundering the intellectual commons and used to deskill labor. Your boss wants AI he can run at the firm level so he captures your skills instead of a lab in California.
1
1
1
1
75
Mark Needham retweeted
The idea that open-weight models are analogous to free and open-source software (FOSS) is propaganda.
Replying to @chamath
As a multi-decade developer, I have to point out that this presentation is not forthcoming. The word "open" is not exactly what is going on here. ​ Generally, this "open" entity is just a black box that you can run on your own hardware, and of course, Jensen is more than happy to sell you the very hardware needed to run it. :-) ​Releasing model weights without providing the underlying training data or the actual source code lacks the transparency and reproducibility required of genuine open systems. ​The Open Source Initiative breaks down exactly why this distinction is so critical: opensource.org/ai/open-weigh…
1
3
201
Mark Needham retweeted
Tony Hoare on corporate languages and the price of reliability: "The price of reliability is the pursuit of the utmost simplicity. It is a price which the very rich find most hard to pay."
4
6
48
1,310
The controversy over Chinese open weight models is kayfabe imo. Local models might be marginally preferable, but they're still not tools of the people.
Replying to @mbneedham
Frustrating how many on the left are falling for this. "Open" models are still weapons of capital, created by plundering the intellectual commons and used to deskill labor. Your boss wants AI he can run at the firm level so he captures your skills instead of a lab in California.
1
1
3
60
Look at the signatories! xitter.cf/satyanadella/status/20…
Open-weight models are essential to a healthy AI ecosystem. Together with others across our industry, we are outlining a path for open-weight models to strengthen American competitiveness and expand economic opportunity, while protecting national security. microsoft.com/en-us/corporat…
1
2
34
I wonder... what happens to the copyright liability if the labs go bust?
Anthropic's S1 reveals that the company considers copyright law a massive risk. It says: "If we cannot rely on the fair use doctrine to access copyrighted work for training our models, we may be required to pay significant fees to obtain a license to such training data or our ability to further develop our models [...] and distribute models trained on such data at issue could be adversely affected." Adverse rulings "could require us to retrain affected models from scratch at substantial cost." It notes that the copyright questions around AI training have not been resolved in the US, and that other countries don't even have 'fair use'. There is hope yet.
2
64
Mark Needham retweeted
Whole country susceptible to good cop bad cop.
15
1,072
17
7,798
51,667
Mark Needham retweeted
we could've fought world war ii on american soil and did less damage to our cities
"When I left America at that time, I told a New York reporter that Indianapolis was the most beautiful city in America." - Arthur Conan Doyle, absolutely correct.
4
24
844
13,846
South Arabian Alabaster Anthropomorphic Stele
33
378
114
3,648
219,471
Mark Needham retweeted
Replying to @burkov
IP isn't the only problem. Existing sites of knowledge production (academic publishing, Stack Overflow, etc.) aren't just 'data', they're -systems- with feedback mechanisms for correcting, incentivizing and validating intellectual work. No attribution means those processes break.
1
2
203
there is an argument (which I've made privately until now) that one of the principal innovations of the large language model labs is a *legal* innovation, not a technical one by building a system that hoovers up mind-bogglingly large volumes of human information and feeds it into a recombinatorial association machine, the companies have bulldozed the existing IP regime and made the contents of ~all human IP (including the most mundane) searchable, reconfigurable, and usable while also obscuring any trace of origin and eliminating a need to provide credit or attribution. much of the perceived benefit is therefore access and exposure to IP people would not otherwise have had access to, presented behind a system that abstracts the IP from its context and history and presents it as net-new information. I imagine we would have seen ~similar productivity outcomes had Google actually been allowed to make productive use of the Google Books database over a decade ago.
I’m starting to get increasingly uncomfortable with the economics of AI as it is currently happening It seems fundamentally wrong that a small number of people with access to a huge amounts of VC money can vacuum up all the contributions of humanity, often paying nothing for it, then turn around and get enormously rich while the many people who put all the work into creating the training data get nothing I am generally pro-AI, but the way it is happening is seeming increasingly parasitic to me
72
203
46
2,114
96,226
Enclosure of the intellectual commons.
Replying to @mr_scientism
The incentive for AI companies is to make everything proprietary. All knowledge must be held internally. New knowledge can only be created internally. External sources of knowledge must be destroyed.
1
3
11
219
Replying to @molly0xFFF
My point is that it's not necessarily against the interests of the big players to destroy the commons. Obviously it's bad for us, but they may even *want* to degrade public sources of information. Among other things, it pulls the ladder up on potential competitors.
1
54
Yet again, the Game UI Database is being scraped to the point that it's getting knocked offline. Even Cloudflare's "I'm under attack" mode isn't working. I just wanted to create a resource for real designers. I'm getting so fucking sick of this.
93
338
17
5,268
140,048
Mark Needham retweeted
Why didn't OpenAI make the AI clean up the math papers before publishing them? Because it can't. The AI messes around doing random stuff at an unbelievable rate until it comes across a result. Its level of understanding and quality of guidance is still minimal.
64
68
8
856
31,796
Mark Needham retweeted
A Meta Brasil whistle-blower told reporter Laís Gouveia that it fired and replaced most of its content moderators with AI during election season. "Decisions are now coming from a single automated command, possibly issued in the US, designed to boost some content & bury the rest”+
46
1,176
112
3,285
50,276
I keep coming back to the fact that most overweighting of language model capabilities comes from poor intuition about the underlying data and breadth of human work fueling it (which is understandable; wrapping your head around the breadth of writings and online behaviour of billions of people is not for the faint of heart! but it *is* critical) Millions of subject-matter experts’ work and writing, stripped of its original source and context, gets ingested into a massive association machine and presented as new work without history by a generative text engine. As more people use the private language models and live work-in-progress gets ingested by them, their results become more current to the state-of-the-art, but creates the new problem of active work by human experts getting stolen and presented as machine innovation by the marketing machine (when the goal should have always been to highlight how the machines augment and facilitate human expert work rather than ‘create superintelligence’). So much of the hype also has a Gell-Mann quality about it, ooh-ing and aah-ing about grand announcements outside one’s domain while simultaneously dismissing the critique of domain experts out of a mix of faith in a machine god and desire to knock the experts down a peg. It’s a toxic mix.
Anthropic's discovery of array-associated reverse transcriptases? Turns out it was already done by a Danish scientist. He used Claude for administrative tasks, but not in an agentic manner for the discovery. He already filed a patent and gave talks on it nytimes.com/2026/09/27/scien…
3
18
1
81
4,327
I gotta say, I don't like how often I see people posting Google AI summary results online as supporting evidence.
6
5
1
154
3,729