alfor

16 ms·

151.
by alfor 4y ago
We are sleepwalking into disaster in all the directions. I think this movement will collapse and people will return to traditions/religion. Others part of society that don’t will self extinguish as is already apparent in the birth rate
152.
by alfor 4y ago
Do you know what kind of speed you can expect with inference on CPU?
153.
by alfor 4y ago
Thanks, so to make usable you would need a GPU with at least 50GB. That probably mean A100 with 80GB at around 10K$. Ok I understand why people use CPU and main memory. After a quick check up you can rent a A100/80G at 1$-2$/h.
154.
by alfor 4y ago
What kind of GPU is needed to run the 65B models, can a consumer grade card do the job? Is it worth it? Or do you use cloud instances?
155.
by alfor 4y ago
I will try it, thanks.
156.
by alfor 4y ago
Many are building such things. But I think they are mostly pointless as OpenAI is so far ahead of everyone external it’s not even funny. Most externals things with the API will be obsolete in a few months. They had GPT4 6 months ago or more
157.
by alfor 4y ago
The training is to maximize good answers. Now there is lot of wrong answers that are close to the right one and ChatGPT does not expose it at the moment. But in the API you can see the level of confidence in each world the LLM output.
158.
by alfor 4y ago
Google and others had strong AI, but they are keep internal to maximize profits and spy on people. OpenAI put it out there so we can see it, interact and have the conversation. They have way more inside, ChatGPT is there to test it out in t
159.
by alfor 4y ago
I think it’s mostly different because they crippled the public version for now: no internet access, everything is done in one pass. In our mind we get an idea, we inspect it, we try different variations, we simulate how it will be perceive(
160.
by alfor 4y ago
ChatGPT is human level intelligence, it’s not just novel and cool, it’s the thing. Remember, GPT-4 training was finished 6 months ago. Listen to people at OpenAI, their concern is: disruption to the world, UBI, getting people used to superi
161.
by alfor 4y ago
One of the most frustrating thing is that GPT-4 training was finished 6 months ago. It might as well be a decade in the current movement of things. What that mean: - OpenAI has such a large head start that they are only worried about safety
162.
by alfor 4y ago
I agree with you. I think Apple is trailblazing again. Most laptop will go in that direction. Tighter integration will lead to a better performance/cost and better efficiency also. AMD 3D cache is also going in that direction integrati
163.
by alfor 4y ago
Much better a imperfect plan than no plan. I hope it will help the industry/government to push things faster. It seem to me there is a huge problem of incentives at the moment: the price of electricity is mostly constant (in most place
164.
by alfor 4y ago
It’s not a very meaningful problem now, scaling is. If you have a growth of 50% per year (Tesla) it mean that the 20 years recycling is 1.5*20 = 3325X less. I hear that most of the recycling is from scrap production, if you have a 1% defect
165.
by alfor 4y ago
I thought we were going into chiplets and tighter integrations. A bit like what Apple did with M1/M2, putting ram/flash/cpu/gpu on the same package with very short, wide and fast interconnects. Leaving only IO, specializ
166.
by alfor 4y ago
1- Media machine 2- He took position against the left mass censorship and narrative
167.
by alfor 4y ago
Religion/Christianity Before downvoting hear me out. We are meat based neural network that evolved for certain conditions (small tribes of hunter). Religion was the way to pass knowledge down generations, things that don’t happen often
168.
by alfor 4y ago
In 2023 what is the advantage of mariaDB/mysql vs Postgres?
169.
by alfor 4y ago
It’s because Google is a monopoly it doesn’t operate on normal economic incentives. It’s goal is only to keep the monopoly: appear bening, keep tech advance in house, share it’s spying network with government so the government don’t regulat
170.
by alfor 4y ago
How the two parts of sheet metal are joined together? How are the reinforced inside? Make me think of a design like a zipper, but I don’t know at that scale what would work, magnets? Maybe there is a way to transfer the loads to more than t
171.
by alfor 4y ago
I like this explanation. Lot’s of multitasking, over consumptions of information and procrastination on my side. With med it’s like I am more calm and energized at the same time. Made me see how most people see the world for the first time.
172.
by alfor 4y ago
I don’t see a way out of this with the GPT/AI able to create fake persona in an instant.
173.
by alfor 4y ago
Depend on version of fsd. They used to go through a normalization process that tried to make the video similar in differents conditions. This has been ripped off and now the signal go strait to NN. I heard (and saw video) where the camera d
174.
by alfor 4y ago
Look at the performance/power of the M1/M2, this is a result of tighter integration. I agree with you Apple does a lot of price differentiation with the amount of storage/ram where it does not reflect at all the cost of the c
175.
by alfor 4y ago
ADHD is probably the problem under this behaviour. Can’t complete tasks and underperform but want to help other, be part of the team. People with ADHD have developed coping behaviour to fit in but have very low control on their executive fu
176.
by alfor 4y ago
Completely agree. It’s more wasteful than the path the we are in (full integration) Laptop will more and more look like an Iphone (glue together, with no replaceable parts) Every connector, every wire, every locking mechanism, every screw h
177.
by alfor 4y ago
Camera are better in low light than our eyes Computers are faster than our brain Software (Neural Net) are not at the level of our brain, in particular they do not fine tune constantly to the current conditions and problems. But they are ge
178.
by alfor 4y ago
GPT-4 training was finished last summer. Karpathy is working on a JARVIS there. I think they already have something quite good for internal use, maybe trying to make it "safe" before releasing it. My guess: we will have a competen
179.
by alfor 4y ago
It will be a large problem 20 years later ... in 2043 But there is already companies that do it now, used batteries are basically very high grade ore, well worth to recycle. But there is also a good change that we will have better chemistry
180.
by alfor 4y ago
Yes but they have huge system of that kind that they use internally on user data. We are all spied on and Google is the biggest player. The government let them get away with their monopoly on the market as long as they have a hand inside of

More ›