Links for 2022-12-26
How many men are raising a child that is not their own? What the genetic data show us: “A common urban myth is that many fathers are cuckolded into raising children that genetically are not their own. Human EPP rates have stayed near constant at around 1% across several human societies over the past several hundred years.” https://threadreaderapp.com/thread/1606294321799475200.html
"The simple fact of the matter is that not a single scientific study has demonstrated that beliefs in conspiracy theories and misinformation have, generally, increased over time." https://knightcolumbia.org/content/whats-the-harm
“We Evaluated ChatGPT vs. Google on 500 Search Queries” https://www.surgehq.ai/blog/googles-existential-threat-chatgpt-matches-googles-performance-on-informational-search-queries-and-smashes-it-on-coding
“At OpenAI a lot of our work aims to align language models like ChatGPT with human preferences. But this could become much harder once models can act coherently over long timeframes and exploit human fallibility to get more reward.” https://threadreaderapp.com/thread/1603862969276051457.html
High-level hopes for AI alignment https://www.lesswrong.com/posts/7BWmLhFtqzqEPs8d5/high-level-hopes-for-ai-alignment
AI alignment is distinct from its near-term applications https://ai-alignment.com/ai-alignment-is-distinct-from-its-near-term-applications-81300500ad2e
List of introductory overviews of AGI safety https://docs.google.com/document/d/1zx_WpcwuT3Stpx8GJJHcvJLSgv6dLje0eslVKvuk1yQ/edit#
Applied Linear Algebra Lecture Series https://www.lesswrong.com/posts/LKAXgTen4Xbqb8eZY/applied-linear-algebra-lecture-series
Next generation protein-based materials capture and preserve projectiles from supersonic impacts https://scitechdaily.com/ground-breaking-new-shock-absorbing-material-can-stop-supersonic-impacts/
Dynamic and Optical Characterization of Dusty Plasmas for Use as Solar Sails http://www.rbsp.info/rbs/PAPER/STAIF02/index.html
“Project Relate is a Google app that allows users with speech impediments to create a customized voice model so they can interact with Google’s voice assistant. After the user trains the model it can pretty effectively parse what they are saying even when other humans can’t.” https://sites.research.google/relate/
“Announcing a holiday gift: 🎅SantaCoder - a 1.1B multilingual LM for code that outperforms much larger open-source models on both left-to-right generation and infilling!” https://twitter.com/BigCodeProject/status/1605958778330849281
“One should remember that Russia tried to build a million-man professional armed forces for 2 decades and failed miserably. That´s why they still have a general conscription system, and still, they couldn´t reach the million soldier’s target.” https://threadreaderapp.com/thread/1606235198244286464.html
There was an exchange between Alex Epstein and OpenAI CEO Sam Altman about ChatGPT refusing to give arguments in favor of fossil fuels. And I suspect this will strike many people as absurd or irrelevant. Why would you want an AI to lie in favor of fossil fuels? But it's not so easy.
Remember that Earth today is ~1 °C warmer than the pre-industrial baseline. We could have prevented this if we had stopped economic growth in 1850.
But would a world that is 1 °C cooler yet economically and technologically on an 1850-level really be better off than the world we live in? I believe that the answer should be a resounding NO. The use of fossil fuels allowed humanity to reach unprecedented levels of prosperity. It was a fundamental driver of the industrial revolution.
This raises the question, at what point should we have stopped using fossil fuels?
Recently, 850 million Chinese people have been lifted out of extreme poverty. China is responsible for ~60% of the increase in global emissions of CO2 since Greta Thunberg was born. Was this worth it, or should they better have stayed poor? And what about developing countries like India? Should we tell them to slow down, even if curbing economic growth would mean postponing and depriving billions of people of access to revolutionary technological and medical advances?
As you can see, arguing in favor of fossil fuels isn't quite as absurd as you might have believed.
If climate change truly is a catastrophic or existential risk, we need other solutions than the Draconian measures required to dismantle capitalism and rob billions of people of the prosperity they deserve and desire.
I'm confident that Nick Davidov is wrong about this. If he had instead written GPT-5 and 2026, then I wouldn't be so sure, though.
In any case, losing this bet isn't too bad. Because, if Davidov is right, we'll soon all be dead anyway (or, if you are not concerned about alignment, at least live in a world in which money will soon matter much less).