Really incredible job, really exciting to see so many great projects come out of Catalyze. Hopefully people will consider funding not just the projects, but also consider the new incubator which created them!
On a side note, I am especially excited about TamperSec and see their work as the most important technical contribution that can be made to AI governance currently.
MathiasKBšø
Karma: 6,482
Donāt put all of your savings into shady cryptocurrencies. If it sounds good to be true, it is because it probably is.
that when I wrote the comment, the post was at ā4 upvotes!
Why on earth are people downvoting this post?
Figuring out how to respond to the USAID freeze (and then doing it) is probably the most important question in global health and development right now. That there has been virtually no discussion on the forum so far has frankly been quite shocking to me.
Have a fat upvote, wishing you the best of luck
More dakka is to pour more firepower onto a problem. Two examples:
Example: ābright lights donāt help my seasonal depressionā. More dakka: āhave you tried even brighter lights?ā
Example: we brainstormed ten ideas, none of them seem good. More dakka: āTry listing a 100 ideasā
Thank you for pursuing this line of argument, I think the question of legal rights for AI is a really important one. One thought iāve had reading your previous posts about this, is whether it legal rights will matter not only for securing the welfare of AI but also for safeguarding humanity.
I havenāt really thought this fully through, but hereās my line of thinking:
As we are currently on track to create superintelligence and I donāt think we can say anything with much confidence about whether the AI we create will value the same things as us, it might be important to set up mechanisms which make peaceful collaboration with humanity the most attractive option for superintelligent AI(s) to get what they want.
If your best bet for getting what you want involves eliminating all humans, you are a lot more likely to eliminate all humans!
I have pushed the idea on the CE research team to the point Iām sure theyāre sick of hearing me rant about it!
To my knowledge itās on their list of ideas to research for their next round of animal welfare charities
Merry Christmas and happy holidays :)
I havenāt looked into this at all, but the effect of eradication efforts (whether through gene drive or the traditional sterile insect technique) is that screwworm stop reproducing and cease to exist, not that they die anguishing deaths.
Iām in San Francisco this weekend and next week, if youāre in the bay area want to meet, donāt hesitate to reach out :) (both happy to meet 1-1 and at social events)
Absolutely do so! In the eyes of the vast majority of employers, organizing a university group centered around charity shows character and energy, highly positive qualities in an employee.
How big a deal is the congressional commission? What is the historical track record of Congress implementing the commissionās top recommendation?
With hindsight, this comment from Jan Kulveit looks prescient.
I decided to drown a puppy in a local pond. Hopefully, doing so would toughen its character, rather than allowing it to succumb to modern frailty.
Laughed out loud for a good minute after reading this!
I strongly upvoted this post because Iām extremely interested in seeing it get more attention and, hopefully, a potential rebuttal. I think this is extremely important to get to the bottom of!
At first glance your critiques seem pretty damning, but I would have to put a bunch of time into understanding ACEās evaluations first before I would be able to conclude whether I agree your critiques (I can spend a weekend day doing this and writing up my own thoughts in a new post if there is interest).
My expectation is that if I were to do this I would come out feeling less confident than you seem to be. Iām a bit concerned that you havenāt made an attempt at explaining why ACE might have constructed their analyses this way.
But like Iām pretty confused too. Itās hard to think of much justification for the choice of numbers in the āImpact Potential Scoreā and deciding the impact of a book based on the average of all books doesnāt seem like the best way to approach things?
Doesnāt look like it
This is what my frontpage looks like
Naively, is there a case for using the average of the two?
Two ideas off the top of my head
Distribution of charitable goods (such as insecticide nets or cash-transfers) and its effect on economic growth of the region/ācountry
Proportion of people identifying as flexitarian/āvegetarian/āvegan and the effects on sales of plant-based products.
EDIT: Someone on lesswrong linked a great report by Epoch which tries to answer exactly this.
With the release of openAI o1, I want to ask a question Iāve been wondering about for a few months.
Like the chinchilla paper, which estimated the optimal ratio of data to compute, are there any similar estimates for the optimal ratio of compute to spend on inference vs training?In the release they show this chart:
The chart somewhat gets at what I want to know, but doesnāt answer it completely. How much additional inference compute would I need a 1e25 o1-like model to perform as well as a one shotted 1e26?
Additionally, for some x number of queries, what is the optimal ratio of compute to spend on training versus inference? How does that change for different values of x?
Are there any public attempts at estimating this stuff? If so, where can I read about it?
I do a lot of writing at my job, and find myself using AI more and more for drafting. I find it especially helpful when I am stuck.
Like any human assigned with a writing task, Claude cannot magically guess what you want. I find that when I see other people get lackluster writing results with AI, itās very often due to providing almost no context for the AI to work with.
When asking for help with a draft, I will often write out a few paragraphs of thoughts on the draft. For example, if I were brainstorming ideas for a title, I might write out a prompt like:
Perhaps Claude comes up with two good titles, or one title has a word I particularly like. Then I might follow up saying:
From this process, I will usually get out something good, which I wouldnāt have been able to think of myself. Usually Iāll take those sentences, work them into my draft, and continue.