Scimitar
1yr ago
Yes last it heard it was reduced to 25 messages/3 hours, and that they were possibly reducing it further.
Scimitar
1yr ago
What are the current usage limits? Does the history work? What is the maximum input length for a single comment (presume 8k tokens)?
Scimitar
1yr ago
I get good summaries by just pasting the content along with the word "Summarize" or "Summarize the above" at the end.
If you use a more imperative rather than conversational tone, it is more likely to do what you ask, rather than pad out the response with empty patter.
Scimitar
1yr ago
Yeah, just like mobile dev exploded and was lucrative over the last decade, the next few years might be the age of the "applied AI engineer".
Scimitar
1yr ago
The OpenAI team use it internally, so it can't do it by itself but at least it can help!
Scimitar
1yr ago
Screw a filler plate on top the existing latch plate? Something like this. Non-destructive and easy to reverse.
Scimitar
1yr ago
Scott Aaronson describes a feeling (that I too experienced):
Here’s the thing: I spent my formative years—basically, from the age of 12 until my mid-20s—feeling not “entitled,” not “privileged,” but terrified. I was terrified that one of my female classmates would somehow find out that I sexually desired her, and that the instant she did, I would be scorned, laughed at, called a creep and a weirdo, maybe even expelled from school or sent to prison.
I was not as bad as Aaronson, but I held a completely unquestioned conviction that no girl must ever know how much I desired her, and that no one must know I had any sexual thoughts at all. I have no idea where it came from, but it seemed as evidently true to me as the fact that the sky is blue.
So, reading this story, I wonder if this innate impulse is actually adaptive for spergs. If you don't understand the social landscape of romance and dating, then indeed your best bet is to opt out and hide. If you try to play without understanding the rules, you end up ostracized or worse.
Scimitar
1yr ago
I think I broadly agree. As it stands today, many (most?) organisation are horribly inefficient. But what this does is raise the ceiling of maximum productivity. We might be entering a world of hyper-startups, a world where a single guy can write code, write marketing copy, produce art assets, provide customer support, etc, and create more value than a competitor deploying 1000x more resources.
Scimitar
1yr ago
Disagree somewhat.
You go into the hospital with trouble breathing, your doctor comes to see you. Your heart rate is elevated. Do you have a growing infection? Are you nervous talking to the doctor? Were you trying to work out because you have a date next week? Is this a side effect from the breathing medication we gave you? Were you just fucking your girlfriend? One of these requires immediate start of antibiotics, and patients can have more than one of them happening at the same time (and in my experience, have).
Yes, this gestalt reasoning, this gut feeling, "does the patient look sick?" is important, but significantly this is now a thing that machines can do, and are continuously improving at.
Current decision support tools can't even read an EKG
Yes the interpretations that are printed out on those machines are shit. But this is not state of the art. It is possible to do better
We develop an algorithm which exceeds the performance of board certified cardiologists in detecting a wide range of heart arrhythmias from electrocardiograms recorded with a single-lead wearable monitor.... We exceed the average cardiologist performance in both recall (sensitivity) and precision (positive predictive value).
Scimitar
1yr ago
I found an comparison between davinci-003 and chatgpt. Seems like chatgpt is better at some things and worse at others; not a blanket downgrade: https://scale.com/blog/chatgpt-vs-davinci.
Also there is no more data flywheel:
Data submitted through the API is no longer used for service improvements (including model training) unless the organization opts in
How have they cut costs? I've seen answers I don't understand. It doesn't seem like they are necessarily running at a loss:
Quantizing to mixed int8/int4 - 70% hardware reduction and 3x speed increase compared to float16 with essentially no loss in quality.
A*.3/3 = 10% of the cost.
Switch from quadratic to memory efficient attention. 10x-20x increase in batch size.
So we are talking it taking about 1% of the resources and a 10x price reduction - they should be 90% more profitable compared to when they introduced GPT-3.
I did some quick calculation. We know the number of floating point operations per token for inference is approximately twice the number of parameters(175B). Assuming they use 16 bit floating point, and have 50% of peak efficiency, A100 could do 300 trillion flop/s(peak 624[0]). 1 hour of A100 gives openAI $0.002/ktok * (300,000/175/2/1000)ktok/sec * 3600=$6.1 back. Public price per A100 is $2.25 for one year reservation.
https://news.ycombinator.com/item?id=34986915
Essentially, I had previously thought that better language models will need to be orders of magnitude larger and orders of magnitude more expensive. Now it seems clearer that we're moving towards a future of intelligence too cheap to meter. And "endless stupidity generator"? I just plain disagree, and feel compelled to build a thin client product on top of the API to prove it. There are many jobs that are mostly low-skill word manipulation, and now that you might get that 100x or 1000x cheaper, it opens up yet unimagined opportunities IMO.
Scimitar
1yr ago
A few months ago OpenAI dropped their API price, from $0.06/1000 tokens for their best model, to $0.02/1000 tokens. This week, the company released their ChatGPT API which uses their "gpt-3.5-turbo" model, apparently the best one yet, for the price of $0.002/1000 tokens. Yes, an order of magnitude cheaper. I don't quite understand the pricing, and OpenAI themselves say: "Because gpt-3.5-turbo performs at a similar capability to text-davinci-003 but at 10% the price per token, we recommend gpt-3.5-turbo for most use cases." In less than a year, the OpenAI models have not only improved, but become 30 times cheaper. What does this mean?
A human thinks at roughly 800 words per minute. We could debate this all day, but it won’t really effect the math. A word is about 1.33 tokens. This means that a human, working diligently 40 hour weeks for a year, fully engaged, could produce about: 52 * 40 * 60 * 800 * 1.33 = 132 million tokens per year of thought. This would cost $264 out of ChatGPT.
https://old.reddit.com/r/singularity/comments/11fn0td/the_implications_of_chatgpts_api_cost/
...or about $0.13 per hour. Yes technically it overlooks the fact that OpenAI charge for both input and output tokens, but this is still cheap and the line is trending downwards.
Full time minimum wage is ~$20k/year. GPT-3.5-turbo is 100x cheaper and vastly outperforms the average minimum wage worker at certain tasks. I dunno, this just feels crazy. And no, I wont apologize for AI posting. It is simply the most interesting thing happening right now.
Scimitar
1yr ago
Marginal revolution have free online courses, but I haven't taken them and can't vouch for them
Scimitar
1yr ago
Chatgpt is great for a certain class of question where you know the answer is true once you see it. E.g. ask it how to do something with a certain JavaScript library and it will give you the code. It's then very easy to confirm the validity by cross checking the documentation.
Generally I treat chatgpt as a wise but crazy old man. It will say insightful things, give you reference and names and ideas, but it's up to you to go away and confirm and research and come back with followup questions. Using chatgpt has to be embedded with your own research process. It can't be used in isolation, like you say.
Scimitar
1yr ago
97th percentile means out of a room of 33 random people, you're the smartest. Does that feel right to you? In my school year of 200, there was one kid who was clearly the brightest. I could believe he was a 1 in 1000 intelligence = ~145 IQ. I think I was roughly second, so smartest 1 in 100 or 135 IQ. Yet this was a good school full of middle class kids, so the selection isn't fair. The hard part about building this intuition is getting an idea of what a room full of truly random people feels like.
(Also fixing broken machinery as an IQ test would bias against women, who have poorer spacial reasoning as a group. I think women outperform men on verbal reasoning, so you have to add some of that in.)
Scimitar
1yr ago
I liked the first season, but I think I stopped watching in the middle of season 3. Felt like the characters became caricatures, and that it became less about startups and more about interpersonal drama. Loved the 80s aesthetic though.
Scimitar
1yr ago
Did the greeks/romans/ancients write about leadership and management? (surely they did) Any specific recommendations here? I have not read much at all from the classical world, other than Seneca's letters, which are mostly about personal matters.
Scimitar
1yr ago
So perhaps the point of Plato's cave is to pose the question of whether the prisoners are not just only experiencing a 'flatter' version of reality, but whether they are fundamentally less able to perceive things in 'higher' reality.
IMO it's not so much that the cave-dweller is physiologically unable to perceive Truth, in the way that those cats can't see horizontal lines, but about the mind-constraining effects of socialisation/indoctrination/brainwashing that is an inevitable and necessary result of growing up inside a culture.
Once the cave-dweller is shown the outside world/Truth, he recoils. We like to imagine that upon seeing Truth, we will not only recognise it, but accept it. But Plato is right. The cave is comfy and familiar, and besides, all our friends and family are there. And once the ex-cave-dweller accepts the Truth, and returns to the cave, he finds himself mocked and ostracised.
Those who take the red pill often don't end up saving the world. They might end up poor and mad, like Ignaz Semmelweis or van Gogh.
Scimitar
1yr ago
Haha I guess it is a contrarian take. Most/all of his content isn't political, not in the way that people like John Oliver, John Stewart, Glenn Beck, etc, are. If you watch World Peace, the sketches are things like a parody of a /fit/ guy, or a wife who cucks her husband with an IVF sperm donor, or a pickup artist helping a disabled guy get laid.
I mean, a lot of the comedy is subversive. It's funny precisely because you can't say it, e.g. MDE released a book titled 'How to bomb the US government'. So in a world where you're not supposed to quote crime statistics, standing up in front of a crowd and quoting crime statistics becomes funny, but also gets you labelled as all kinds of things. Or, after getting your show on adult swim, wouldn't it be funny to sing "Jews rock!", and include footage of producers looking uncomfortable?
I don't see Sam as a culture warrior, like Nick Fuentes or Milo. During the BLM protests a couple years ago, he was telling people not to protest/counter-protest, not to get arrested, it's not worth it, just be cool, etc. I feel like he's generally sceptical of politics, he doesn't have an agenda, he's not trying to enact change. He has anti-establishment views, and satirizes the PMC, the Davos elite, etc, but I feel like it comes from a position of accepted powerlessness.
As for 'virulent racist', that's just not the vibe I get after watching him. I mean, Sam invited Harley Morenstein (Epic Meal Time guy, and jew) to come and train with him. Sam made holocaust denial jokes and they both laughed. It feels more like a "I don't care what race you are as long as you're racist" type thing.
Scimitar
1yr ago
Yes, he is a comedian first and foremost. If he thinks something is funny, he'll say/do it. That's the main driving force. Most of the time he is mocking people he believes deserve to be mocked, or putting people into uncomfortable situations. He will make racist jokes. I don't believe him to actually be a virulent racist or antisemite or whatever. I don't even think he is a particularly political person.
Is he smart?
98th percentile sounds right. I think he's currently running his own company with a few employees. He is not a moron.
Is he a sociopath?
No, not for my understanding of 'sociopath'. But he is willing to commit to the bit and put on a production, like what he did to idubz. Again, in the name of comedy, and in that specific case, because Sam thought idubz was going to negatively portray him.
Scimitar
1yr ago
This is a long shot. I am trying to track down the source of a half-remembered quote I saw on twitter, I believe a screenshot from a book. It went something like this.
Imagine a man who devotes himself to potatoes. He lives his entire life in a room made of potatoes, he sleeps in bed made of potatoes, he eats only potatoes, etc. Now imagine another man who lives a more varied life, who knows a potato, but also knows turnips, onions, cabbage, etc. Which of these men has a greater understanding of the nature of potatoes?
The point being that the first man has no point of reference. Although he has spent far longer with potatoes, he lacks the framing and context that the generalist has. Appreciate any leads on this one!
Scimitar
1yr ago
Not sure what you mean but tech companies have all got their own in-house GPT-3 equivalents.
Scimitar
1yr ago
Are you reading it right? The numbers in parens are the differences.
E.g. looking at 'The primary reason I interact with Aella's web presence is a (dimly) conscious desire to do something sexual with her.'
The male average is -0.33 (between neutral and somewhat disagree)
The female average is -2.21 (disagree)
The 'bang a hot 15yo' is -0.5 male, -2 female
So both of these questions are disagree, but women disagree harder.
What's your definition of AGI? The label feels more like a vibe than anything else.
For me, the multi-modal capabilities of GPT-4 and others [1][2] start to push it over the edge.
One possible threshold are Bongard problems[3]. A year ago I thought, while GPT-3 was very impressive, we were still a long way from AI solving a puzzle like this (what rule defines the two groups?) [4]. But now it seems GPT-4 has a good shot, and if not 4, then perhaps 4.5. As far as I know, no one has actually tried this yet.
So what other vibe checks are there? Wikipedia offers some ideas[5]
Turing test - GPT3 passes this IMO
Coffee test - can it enter an unknown house and make a coffee? Palm-E[1] is getting there
Student test - can it pass classes and get a degree? Yes, if the GPT-4 paper is to believed
Yes, current models can't really 'learn' after training, they can't see outside their context window, they have no memory... but these issues don't seem to be holding them back.
Maybe you want your AGIs to have 'agency' or 'conciousness'? I'd prefer mine didn't, for safety reasons, but would guess you could simulate it by continuously/recursively prompting GPT to generate a train of thought.
[1] https://ai.googleblog.com/2023/03/palm-e-embodied-multimodal-language.html
[2] https://arxiv.org/pdf/2302.14045.pdf
[3] https://metarationality.com/bongard-meta-rationality
[4] https://metarationality.com/images/metarationality/bp199.gif
[5] https://en.wikipedia.org/wiki/Artificial_general_intelligence#Tests_for_testing_human-level_AGI
More options
Context Copy link