ai
I've been reading up on the Lottery Ticket Hypothesis, which is super interesting.
Basically, the observation is that these days we build vast neural networks with billions of parameters, but most of the parameters aren't needed. That is, after training, you can just throw away 95% of the network (pruning), and it will still work fine.
The LTH paper is asking: could we start with a network just 5% of the size, and get comparable results? If so, that would be a huge performance win for Deep Learning.
What's interesting is that you can do this, but only by training the full network (perhaps several times) to see which neurons are needed. They argue that training a neural network isn't so much creating a model, as finding a lucky sub-network (a lottery ticket) from the randomly initialized network, a bit like a sculpter "finding" the bust hidden in a block of marble.
Initial LTH paper: http://arxiv.org/abs/1803.03635
Follow-up with major clarifications: http://arxiv.org/abs/1905.01067
"Wait, isn't this the plot to the "Terminator" movies?"
https://futurism.com/openai-signs-deal-us-government-nuclear-weapon-security
#news #technology #TechNews #AI #OpenAI #military #terminator
#AI #GenAI #GenerativeAI #LLM #waste #environment
The #EU bans some #AI uses and regulates most of the rest, in service of its citizens and not the "techbros".
AI systems with 'unacceptable risk' are now banned in the EU | TechCrunch
https://techcrunch.com/2025/02/02/ai-systems-with-unacceptable-risk-are-now-banned-in-the-eu/
Resistance to the coup is the defense of the human against the digital and the democratic against the oligarchic.
Defense of the human against the digital has been my mission for some time. Resisting the narratives about how #LLMs "reason", "pass the Turing test", "diagnose illnesses", are "better than humans" in various ways are part of it. Resisting the false narrative that we're on the verge of discovering #AGI is part of it. Allowing these false stories to persist and spread means succumbing to very dark anti-human forces. We're seeing some of the consequences now, and we're seeing how far this might go.
#USPol #AI #GenAI #GenerativeAI #LLM #AGI
Some argue that ai technology is more significant than electricity or the internet, and so it will spread fast. But there is little sign of this. Only 5-6% of American businesses said they used ai to produce goods and services in 2024, according to the country’s Census Bureau.
What's the puzzle? It has its uses but generally this technology is not particularly useful for most people. Countless billions of dollars have been spent hyping it up to pretend that isn't the case. But reality matters.
Nowadays I think of generative AI as a power grab. It's not a useful tool for a lot of people, but it's exceptionally useful to people who want to grab and hold power, or restructure power in their favor.
#AI #GenAI #GenerativeAI
"The original video showing off Google's Gemini offerings in Workspace was posted five days ago and showed the AI generating a response for a small-business owner in Wisconsin. In the response, Gemini claims Gouda accounts for "50 to 60% of the world's cheese consumption.
The edited ad now shows a Gemini response that says Gouda "is one of the most popular cheeses in the world."
https://www.businessinsider.com/google-super-bowl-ad-gemini-ai-gouda-cheese-2025-2
Archived link: https://archive.is/2cX3i
Not sure what the chances of this bill passing are, but it's certainly interesting to see the efforts.
"Under Hawley’s proposed law, “technology or intellectual property” developed in China would be barred from importation into the U.S. Anyone found violating these restrictions could face up to 20 years in prison, as well as substantial monetary penalties of up to $1 million for individuals and $100 million for companies."
https://interestingengineering.com/culture/law-proposed-to-criminalize-deepseek
Unbelievable
#ElonMusk’s US #DOGE Service are feeding sensitive data into #AI software via #Microsoft’s #cloud
#Musk’s US #DOGE Service have fed sensitive data from across the #Education Dept into #ArtificialIntelligence software to probe the agency’s programs & spending…. The AI probe includes data w/personally identifiable info for people who manage grants, & sensitive internal financial data…
#law #security #InfoSec #CyberSecurity #NationalSecurity #Trump #TrumpCoup
https://www.washingtonpost.com/nation/2025/02/06/elon-musk-doge-ai-department-education/
Happy #FishFriday !
Time to #MakeSomethingFishy
**not affiliated with #HashtagGames** but my shameless promotion of my Ebay store where I sell hand crafted aquarium ornaments and other things. But feel free to play along!
https://www.ebay.com/usr/crystal_fish_caves
#handmade #crafts #fedigiftshop #FishofMastodon #pets #diy no #ai photoshop only
"Australia’s Department of the Treasury has found that Microsoft’s Copilot can easily deliver return on investment, but staff exposed to the AI assistant came away from the experience less confident it will help them at work."
https://www.theregister.com/2025/02/12/australian_treasury_copilot_pilot_assessment/
"The New York Times has reportedly approved artificial intelligence tools that newsroom staff can use for editing copy, summarizing information, coding, and writing. The publication announced in an internal email that product and editorial staff will receive AI training, according to Semafor, and introduced a new internal AI tool called Echo for summarizing articles, briefings, and other company activity."
https://www.theverge.com/news/613989/new-york-times-internal-ai-tools-echo
DeepSeek Open Source FlashMLA – MLA Decoding Kernel for Hopper GPUs — https://github.com/deepseek-ai/FlashMLA
#HackerNews #DeepSeek #Open #Source #FlashMLA #MLA #Decoding #Kernel #for #Hopper #GPUs #GitHub #OpenSource #AI
Even though #Signal *itself* has sound security - taken in isolation - it's always moored in an ecosystem (that of smartphones which run #iOS or #Android) which, by default, have #AI set up to shoulder-surf Signal, sending reports back to the mothership. In #iOS, that AI is called "#Apple Intelligence", and in #Android, it's called #Google Assistant. Repeat, these are on by default. *Only a small percentage of your family, friends, colleagues, and fellow country-persons will pain-stakingly disable these.*
Sure, Signal itself is secure by default, but "a chain is only as strong as its weakest link". And the weakest link is the ecosystem which Signal is moored in - that of smartphones policed and patrolled by AI, which report back to their respective motherships *in a strong majority of cases*.
Alternatives like #Deltachat (for normies willing to open their wallets/purses to rent an auto-crypt-compatible email address) and #XMPP (easily free to use, however realistic and mature only for non-iOS/#MacOS-users at present) have comparable E2E encryption. Deltachat and XMPP *don't* require smart-phone "moorings" - thereby making it much more realistic to dodge the almost-pervasively AI-patrolled ecosystems.
New Open at Intel Podcast! I spoke with Andrew Brown of Exam Pro about his free generative AI bootcamp for developers, #Deepseek, keeping up with AI development and its rapidly moving pace, and a lot more. Check it out!
Episode: https://openatintel.podbean.com/e/mastering-generative-ai/
Clip: https://youtube.com/shorts/ofsiK5cF1u0?feature=share
"CEO Marc Benioff said Salesforce, San Francisco’s largest private employer, does not plan to hire engineers this year because of the success of AI agents created and used by the company."
https://sfstandard.com/2025/02/27/salesforce-marcbenioff-layoffs-tech-agents/
#news #TechNews #technology #AI #salesforce #labor #WorkersRights
"Rita Contreras, a hairstylist in Brooklyn, said that in the past six months, the number of clients walking in with AI images has ballooned.
[...]
“I have to just say, ‘This is not a real person. This is not real hair,’”"
https://www.washingtonpost.com/technology/2025/02/23/ai-hair-inspo-wedding-stylists-businesses/
Archived link: https://archive.is/erYG5
I was listening to a podcast where the guest argued: when it comes to art, today's AI is like a tool, but in the future, maybe it could be an artist.
As he explained why, it became clear why I disagree with him. He defines "art" as a product that is valued by art critics and by consumers. I agree that AI generated content may become more common and broadly accepted in the future, but I don't think that's a good thing, and I wouldn't call it art!
I see art as more of an activity than a product. An artist does art. Whatever the output of that process is, valuable or not, is also called "art."
Am I saying only humans can make art? Not at all. But to be "art" it has to be self-expression. The artist must have something to say. As long as we are designing the AI, training it, and prompting it, then it will be a tool. Without us it has no initiative, no opinion, and no creative urge.
#AI #GenAI #GenerativeAI #AGI #Pygmalion #Golem #Pinocchio #Talos #Frankenstein
Meta is facing an AI copyright lawsuit in France brought by authors and publishers who are accusing the company of economic “parasitism.” Read more at @Techcrunch. #Meta #Instagram #Facebook #AI #Copyright #Tech #Technology https://flip.it/mSO6-8
How ProPublica Uses AI Responsibly in Its Investigations
—
When our reporters prompted a large language model to help identify “woke” themes in a database of grants, AI helped them tell a vital accountability story about science funding and Ted Cruz.
https://www.propublica.org/article/using-ai-responsibly-for-reporting?utm_source=mastodon&utm_medium=social&utm_campaign=mastodon-post
#News #AI #Tech #Journalism #Media #Research #Data #Technology