Links for 2023-11-07
GPTs: Users can program agents through a casual dialogue, and GPT asks clarification questions. The agent can use all kinds of plugins, the browser, local files, and a code interpreter. GPTs will continue to get more useful and smarter, and you’ll eventually be able to let them take on real tasks in the real world. https://openai.com/blog/introducing-gpts
GPT-4 Turbo supports 128K context and has fresher knowledge than GPT-4. Its input and output tokens are respectively 3× and 2× less expensive than GPT-4. The platform is becoming multimodal. GPT-4 Turbo with Vision, DALL·E 3, and text-to-speech are all now available to developers. https://openai.com/blog/new-models-and-developer-products-announced-at-devday
A first-of-its-kind study on autonomous vehicles' safety. Published as a result of research collaboration between Swiss Re and Waymo. AVs *much* safer and easier to insure than humans: At least 300% safer than human drivers for bodily damage. At least 400% safer than humans for property damage. https://www.swissre.com/reinsurance/property-and-casualty/solutions/automotive-solutions/study-autonomous-vehicles-safety-collaboration-with-waymo.html
Nvidia Is Piloting a Generative AI for its Engineers: ChipNeMo summarizes bug reports, gives advice, and writes design-tool scripts https://spectrum.ieee.org/ai-for-engineering
"The Chinese government sank billions of dollars into two failed efforts to build foundries for advanced chips; Chinese chatbots have struggled to compete with ChatGPT, because the Party imposed rules requiring them to uphold “socialist core values." https://www.newyorker.com/magazine/2023/10/30/chinas-age-of-malaise [https://archive.ph/oSrFb]
LMs and humans show strikingly similar patterns in how the content of a logic problem affects their answers. https://arxiv.org/abs/2207.07051
Grokking, memorization, and generalization — a discussion https://www.lesswrong.com/posts/BYwGEBspGgPY5nBZN/grokking-memorization-and-generalization-a-discussion
Unleashing the Power of Pre-trained Language Models for Offline Reinforcement Learning https://lamo2023.github.io/
Paul Christiano - Preventing an AI Takeover https://www.youtube.com/watch?v=9AAhTLa0dT0
“Why Conscious AI Is a Bad, Bad Idea” https://nautil.us/why-conscious-ai-is-a-bad-bad-idea-302937/
Political links:
“I don’t want to live in a country with Hamas supporters. Simple.” — Douglas Murray https://www.youtube.com/watch?v=DqzdnPjPhW0
Footage from Sderot on Oct 7. Hamas terrorists approach the car of Odaya Swisa, a mother and her two children, Romi, age 6, and Lia, age 3. The terrorists murder Odaya, but don't see the little girls who were lying on the floor of the car. When police arrive, little Romi is terrified it's the terrorists returning to kill them. When she finds out they are Israeli, she yells "Take us.” https://twitter.com/AvivaKlompas/status/1721410037363462399
China routinely illegally plunders the national waters of other countries and destroys native habitats and fisheries. https://www.smh.com.au/interactive/2023/fishing-rights/
Aftermath of the Ukrainian Storm Shadow/SCALP cruise missile strikes on the Russian Project 22800 Karakurt Corvette Askold.
This is one of the ships that would have been used to attack Ukrainian energy infrastructure this winter.
In addition to 8 Kalibr cruise missile launchers, this ship was armed with its own air defense system - the Pantsir-M air defense system and an Orlan-10 UAV complex for reconnaissance.
Video of the strikes: https://twitter.com/GloOouD/status/1721603014299943099
More Ukraine links:
The Battle of Avdiivka 2023. Historic footage from Ukraine's 47th Mechanized. https://twitter.com/IAPonomarenko/status/1721624222047314060
“A great memo clarifying the facts about US aid to Ukraine, expertly debunking myths with a direct approach.” https://twitter.com/Tatarigami_UA/status/1721740523524760061
Videos Show Ukrainian Special Forces ‘Cleaning Up’ Wagner Fighters in Sudan https://www.kyivpost.com/post/23722