Founder & President @ https://teapawt.net/t.co/iQKGHWaZxC | Creating the Internet of Data
Boston
Joined March 2008
- Tweets6.9K
- Following2K
- Followers2.3K
- Likes5.8K
Aleks Jakulin retweeted
Yesterday @Reuters reported Anthropic has a secret biology lab. Anthropic wouldn't give the address.
I found it in five seconds through an EPA database.
750 Gateway Blvd, South San Francisco; a 441,000 SF life science building where SmartLabs leased space designed for ISO 5 cleanrooms before vacating in 2024. A heavy rigging company pulled a permit at the same address on September 11; one week before the story dropped.
29 sourced hyperlinks.
russwilcoxdata.substack.com/…
@JLDastin
@MikeNelson
Aleks Jakulin retweeted
While all y’all are freaking out about AI systems hacking computers, I’m actually working to make sure that we can formally verify the security of computer systems, so we don’t have to worry about them ever being hacked again. If you are worried about risks, then work to build practical solutions to mitigate or eliminate them instead of complaining and wringing your hands constantly.
The best way to react to the bicycle danger isn't banning bicycles, or slowing bicycles down.
It's to build roads, lanes, sidewalks, traffic lights - the infrastructure.
I talk more about this in my article linked underneath.
In 1973, a chart in Scientific American showed something delightful:
Put a human on a bicycle, and the same ordinary body becomes one of the most energy-efficient travellers on Earth.
That is what powerful tools do. They amplify us.
AI is becoming a bicycle for the mind. But most AI alignment still happens at the level of one rider: can the model understand what one person wants and help them achieve it?
That is necessary. It is also only the first rung.
A bicycle amplifies a person.
A tandem amplifies a team.
A street amplifies a community.
A street grid amplifies a town.
A road network amplifies a country.
Each new scale requires infrastructure that the one below did not: shared memory, coordination, signals, rules, and ways for people with different destinations to travel together.
A smart model is not yet a smart team.
A smart team is not yet a smart community.
And a planet full of intelligent machines is not automatically an intelligent planet.
The next great AI breakthrough may not be a smarter bicycle.
It may be the road.
1oio.substack.com/p/ai-bicyc…
Everyone calls the new AI results a sign of AGI.
It's fire. Some people have figured out how to hold it and they're roasting pigs. Everyone else is still chewing raw meat and calling the smell hype.
Fire never had a thought in its life. It just made us better at having them.
"Computer Science Achievement and Writing Skills Predict Vibe Coding Proficiency"
arxiv.org/pdf/2603.14133
Aleks Jakulin retweeted
Vredno pozornosti.
Several friends have asked me for my recent media appearances. Now I added them to my website. In particular, there's the discussion on how our society is coping with the advance of artificial intelligence from @BledStratForum, and there's an interview with me at the @SBC_podjetniki
jakul.in/#media
Several friends have asked me for my recent media appearances. Now I added them to my website. In particular, there's the discussion on how our society is coping with the advance of artificial intelligence from @BledStratForum, and there's an interview with me at the @SBC_podjetniki
jakul.in/#media
Every technology that lowered the cost of producing text produced a flood of bad text. Printing press. Mimeograph. Desktop publishing. Blogs. SEO.
We never solved any of them by asking what tool was used. We solved them by getting better at ignoring things.
I spoke of this and more at @BledStratForum earlier this week linkedin.com/feed/update/urn…
Most AI models will do watermarking anyway, e.g. anthropic.com/news/claude-te…
AI use, yes or no, is a false dichotomy. Most people aren't using it correctly, and that's normal: it's a new technology.
We should be thinking about how to use it correctly. There are a few who use it well, and they'll create the future.
Many of the "AI users" correspond to the "internet users" of the past:
teapawt.net/emollick/status/209005…
Aleks Jakulin retweeted
Caka nas najmocnejsi El Nino v merjeni zgodovini. Presegli smo El Nino v 1997-98 in temperatura morja v Pacifiku strmo raste, verjetno gremo prek +4.
Ekonomski vpliv?
Ze bistveno sibkejsi El Nini (97-98) so povzrocili padec svetovnega GDP za 1.6-1.8%.
science.org/doi/10.1126/scie…
AI detection is like word processor detection, typewriter detection, or use-of-Google detection: a knee-jerk reaction to a change in how we do intellectual work.
It might make people feel good, but it doesn't work, and will be seen as a idiotic in the future.
If it expands, it will make things actively worse, just like how spam filtering led to the centralization of the internet.