AI:
We really believe in a path from where we are to AGI and beyond.
Constellation Energy to restart Three Mile Island nuclear plant, sell the power to Microsoft for AI https://www.cnbc.com/2024/09/20/constellation-energy-to-restart-three-mile-island-and-sell-the-power-to-microsoft.html
AI tool led to a striking 26% drop in the number of unexpected deaths among hospitalized patients at St. Michael's Hospital in downtown Toronto. https://www.cbc.ca/news/health/ai-health-care-1.7322671
AI model can reveal the structures of crystalline materials https://news.mit.edu/2024/ai-model-can-reveal-crystalline-materials-structures-0919
RNA language models predict mutations that improve RNA function https://www.biorxiv.org/content/10.1101/2024.04.05.588317v2
Training Language Models to Self-Correct via Reinforcement Learning https://arxiv.org/abs/2409.12917
Iteration of Thought: Leveraging Inner Dialogue for Autonomous Large Language Model Reasoning https://arxiv.org/abs/2409.12618
Making Large Language Models into World Models with Precondition and Effect Knowledge https://arxiv.org/abs/2409.12278
Journey to superhuman performance on scientific tasks: “Here we show experiments that led to the design of PaperQA2, FutureHouse’s scientific RAG system, which exceeds scientist’s performance on tasks like answering challenging scientific questions, writing review articles, and detecting contradictions from the literature” https://www.futurehouse.org/research-announcements/engineering-blog-journey-to-superhuman-performance-on-scientific-tasks
Qwen2.5-Math Demo. Qwen2.5-Math-1.5B/7B/72B-Instruct achieves 79.7, 85.3, and 87.8 respectively on MATH benchmark https://huggingface.co/spaces/Qwen/Qwen2.5-Math-Demo
“The 2024 ARC-AGI competition has passed its midpoint: 3 months down, 2 months to go. The current high score is at 46%. We're announcing new prizes: +100k for the grand prize (now $600k); +$15k for the best paper prize (now $75k)” https://arcprize.org/blog/3-month-update
o1-preview is pretty good at doing ML on an unknown dataset — “In general o1-preview seems like a big step change in its ability to reliably do hard tasks like this without any advanced scaffolding or prompting to make it work.” https://www.lesswrong.com/posts/vLB4tkLWYmQrJstCA/o1-preview-is-pretty-good-at-doing-ml-on-an-unknown-dataset
“Don’t teach. Incentivize: We can’t enumerate every single skill we want from an AGI system because there are just too many of them. In my view, the only feasible way is to incentivize the model such that general skills emerge.” https://www.youtube.com/watch?v=kYWUEV_e2ss
Language Models Learn to Mislead Humans via RLHF https://arxiv.org/abs/2409.12822
How many letters R are there in the word STRAWBERRY? https://fastml.com/how-many-letters-r-in-the-word-strawberry/
Sam Altman says AI reasoning is still at the GPT-2 stage but the improvement curve is steep and the new o1 model represents a new paradigm of AI development which will enable rapid progress in capabilities https://www.youtube.com/live/r-xmUM5y0LQ?si=ZtBKGnmRAOTuN00h&t=3158
OpenAI CEO Sam Altman confirms that Level-3 Agents are coming soon https://www.youtube.com/live/r-xmUM5y0LQ?si=pG-zPMG6sSzr2z9w&t=3254
Jim Fan says that NVIDIA is building a full-stack computing platform for humanoids and other robots, which includes their inference chip, robotics foundation model, and simulation tools. https://www.youtube.com/watch?v=yMGGpMyW_vw
Jensen Huang says we will soon be working with digital agents and people growing up today will have their own personal R2-D2 or C-3PO to follow them through their lives and learn with them https://www.youtube.com/live/r-xmUM5y0LQ?si=TPdr6RPLUT6-Xzvt&t=5115
Ray Kurzweil says the cost of computation has fallen such that half a trillion calculations per second now costs $1, a 75 quadrillion-fold increase since 1931 https://www.youtube.com/watch?v=vuYKvMH3ai4
“We believe AI agents will be part of your team on a daily basis.” https://www.youtube.com/watch?v=48a_aPxucwk
Andrew Ng says Meta used an existing AI model to train their new model by generating synthetic data in an agentic workflow, in an example of using AI to train the next generation of AI https://www.youtube.com/watch?v=0EmzuLylr7I
“Not only was OpenAI's current fundraising round oversubscribed by a staggering amount, The Information is now reporting that for anyone who was lucky enough to make it in OpenAI set a minimum contribution: a quarter of a billion dollars.” https://x.com/AndrewCurran_/status/1836905479232180424
Kling release next version 1.5 of its model, that supports generating 1080p HD video generation. https://www.klingai.com/release-notes
New Paper reCAPTCHAv2 solves 100% of the CAPTCHAs surpassing previous success rates for around 68% to 71%. https://arxiv.org/abs/2409.08831
Neuroscience:
The neurobiology of sentence production: A narrative review and meta-analysis https://osf.io/preprints/psyarxiv/xku24
“Kernel's second-generation Flow2 Time-Domain Functional Near-Infrared Spectroscopy (TD-fNIRS) system meets the requirements of noninvasive and scalable neuroimaging, and uses a validated modality to measure brain function.” https://www.biorxiv.org/content/10.1101/2024.03.12.584660v1
Miscellaneous:
Some plant cells package up messenger RNAs, wrap them in lipid bubbles & deliver them to invading fungal pathogens. The pathogen translates the plant mRNAs into proteins that harm the pathogen. In other words, plants use "information warfare," via mRNAs, to quell invaders. https://www.quantamagazine.org/cells-across-the-tree-of-life-exchange-text-messages-using-rna-20240916/
Columbia Engineers develop new powerful battery "fuel" -- an electrolyte that not only lasts longer but is also cheaper to produce. https://www.engineering.columbia.edu/about/news/new-battery-technology-could-boost-renewable-energy-storage
Mysteries of the Bizarre ‘Pseudogap’ in Quantum Physics Finally Untangled https://www.simonsfoundation.org/2024/09/19/mysteries-of-the-bizarre-pseudogap-in-quantum-physics-finally-untangled/
Politics:
Russia tried to stage coup in Armenia, prosecutors allege https://www.politico.eu/article/russia-stage-violent-coup-armenia-investigation-allegiation/
111 Former G.O.P. Officials Back Harris, Calling Trump ‘Unfit to Serve’ https://www.nytimes.com/2024/09/18/us/politics/republican-officials-harris-endorsement.html [no paywall: https://archive.is/DTJq0]
Ukraine:
“🔥RUSSIAN LOSSES | Luhansk Region🔥 The 🇺🇦Ukrainian 4th Mech Battalion, Presidential Brigade shows huge 🇷🇺Russian losses in the attempts to attack the small village Makiivka, Luhansk region. 🪖🐽x41 invader x3 BMP” https://t.me/JusuMakonis/4043
“Sabotage operations within Russian territory seem to have them so rattled that they’ve resorted to releasing public service announcements, threatening teenagers with 10-year prison sentences.” https://x.com/wartranslated/status/1836361761626243391
“The village of Glushkovo - the bald genocidal grandpa wanted to create a sanitary zone on the territory of Ukraine, now the sanitary zone is the Kursk region.” https://x.com/wartranslated/status/1837142536898384023
Revealed: Russia anticipated Kursk incursion months in advance, seized papers show https://www.theguardian.com/world/2024/sep/20/revealed-russia-anticipated-kursk-incursion-months-in-advance-seized-papers-show
Infantry war at Pokrovsk: why Ukraine's key eastern front started (and continues) to crumble https://www.pravda.com.ua/eng/articles/2024/09/17/7475408/
"Months before President Vladimir Putin’s inauguration in May, he met with Defense Ministry officials who pushed for a fresh round of mobilization to recruit more troops to offset Russia’s losses on the front line in Ukraine, said a person briefed on the exchange. Putin dismissed the idea, saying he wanted to use only those who were voluntarily signing military contracts, the person said...'Forces are currently not sufficient to achieve the original war aims, knock Ukraine out of the war, to undermine its military potential or protect border regions of the Russian territory,' said the person briefed on the exchange with Putin. “'More and more people are saying mobilization is inevitable.'” https://www.wsj.com/world/russia/putin-troops-pressure-russia-ukraine-war-5f242a1c [no paywall: https://archive.is/vP66g]
"Results: The study included 13 649 patient admissions in GIM and 8470 patient admissions in subspecialty units. Non-palliative deaths were significantly lower in the intervention period than the pre-intervention period among patients in GIM (1.6% v. 2.1%; adjusted relative risk [RR] 0.74, 95% confidence interval [CI] 0.55–1.00) but not in the subspecialty cohorts (1.9% v. 2.1%; adjusted RR 0.89, 95% CI 0.63–1.28). Among high-risk patients in GIM for whom the system triggered at least 1 alert, the proportion of non-palliative deaths was 7.1% in the intervention period, compared with 10.3% in the pre-intervention period (adjusted RR 0.69, 95% CI 0.46–1.02), with no meaningful difference in subspecialty cohorts (10.4% v. 10.6%; adjusted RR 0.98, 95% CI 0.60–1.59). In the difference-indifferences analysis, the adjusted relative risk reduction for non-palliative death in GIM was 0.79 (95% CI 0.50–1.24)."
Nothing concrete was found in other words. You cannot just post articles without checking the stats. Takes <2 mins.
https://www.cmaj.ca/content/196/30/E1027