Silly AI, tricks are for kids
This one is going to run long, sorry. This is the text of a Twitter thread on why DeepSeek is scaring the poop out of our domestic AI guys. I found most of it easy to understand. There are a few more tweets in the thread, so if this interests you, check out the original and maybe give the author a follow.
Let me break down why DeepSeek’s AI innovations are blowing people’s minds (and possibly threatening Nvidia’s $2T market cap) in simple terms…
1/ First, some context: Right now, training top AI models is INSANELY expensive. OpenAI, Anthropic, etc. spend $100M+ just on compute. They need massive data centers with thousands of $40K GPUs. It’s like needing a whole power plant to run a factory.
2/ DeepSeek just showed up and said “LOL what if we did this for $5M instead?” And they didn’t just talk – they actually DID it. Their models match or beat GPT-4 and Claude on many tasks. The AI world is (as my teenagers say) shook.
3/ How? They rethought everything from the ground up. Traditional AI is like writing every number with 32 decimal places. DeepSeek was like “what if we just used 8? It’s still accurate enough!” Boom – 75% less memory needed.
4/ Then there’s their “multi-token” system. Normal AI reads like a first-grader: “The… cat… sat…” DeepSeek reads in whole phrases at once. 2x faster, 90% as accurate. When you’re processing billions of words, this MATTERS.
5/ But here’s the really clever bit: They built an “expert system.” Instead of one massive AI trying to know everything (like having one person be a doctor, lawyer, AND engineer), they have specialized experts that only wake up when needed.
6/ Traditional models? All 1.8 trillion parameters active ALL THE TIME. DeepSeek? 671B total but only 37B active at once. It’s like having a huge team but only calling in the experts you actually need for each task.
7/ The results are mind-blowing:
– Training cost: $100M → $5M
– GPUs needed: 100,000 → 2,000
– API costs: 95% cheaper
– Can run on gaming GPUs instead of data center hardware
8/ “But wait,” you might say, “there must be a catch!” That’s the wild part – it’s all open source. Anyone can check their work. The code is public. The technical papers explain everything. It’s not magic, just incredibly clever engineering.
9/ Why does this matter? Because it breaks the model of “only huge tech companies can play in AI.” You don’t need a billion-dollar data center anymore. A few good GPUs might do it.
10/ For Nvidia, this is scary. Their entire business model is built on selling super expensive GPUs with 90% margins. If everyone can suddenly do AI with regular gaming GPUs… well, you see the problem.
11/ And here’s the kicker: DeepSeek did this with a team of <200 people. Meanwhile, Meta has teams where the compensation alone exceeds DeepSeek's entire training budget... and their models aren't as good.
Posted: January 28th, 2025 under personal.
Comments: 6
Comments
Comment from Jon
Time: January 28, 2025, 6:49 pm
Good news: AI will be less expensive.
Bad news: It still sucks.
Comment from Durnedyankee
Time: January 28, 2025, 8:43 pm
Since the entire thing is largely hype from the git go…
It can do anything! It’ll revolutionize the world!
It don’t clean gutters, or unplug drains, buy you groceries.
How often on a daily basis do normal people need to draw pictures, rip through the encyclopedia digging up info on…Malta, and writing a historical synopsis.
Someday, we might have to worry about Colossus The Forbin Project (before the Matrix or Skynet…) But that day is a long way down the road.
Yet, we barely had first news of ChatC-A-T and people were actually mumbling in fear about Skynet.
Very very expensive hype. Talking about the demand on the grid, the need for nuclear power plants to drive them, data centers the size of London.
Infrastructure that will take at least a decade, waiting for designs, and permits, and contractors and construction, and approvals, etc, etc, etc.
Right up there with rich people buying homes on the coast while they warn us about global warming and the rise of sea level…
A way to move money from one bank account to others, or other accounts, into one.
Comment from Some Vegetable
Time: January 28, 2025, 9:57 pm
A couple thoughts…
Edison didn’t invent the light bulb. He just invented the first practical one. Ford didn’t invent mass production; he just utilized it in a field where it provided significant advantage. Quite often the first mover is clumsy while someone who follows finds a better path forward. If we are to have an A.I. future, and we are, I would prefer A.I. to be as cheap and available as possible. I don’t like great power confined to a few, and that has been a worry for me about A.I.
Next, I hesitate to say that I won’t find A.I. useful somehow. Pointless and not very useful is exactly what I said about my family’s first microwave. Can’t cook in it – what good is it? Our first P.C. – there was no internet yet and I wasn’t a gamer. My first cell phone I liked but when I insisted Mrs. Vegetable get one, she carried it around in her purse turned off unless SHE wanted to call someone.
So I don’t want to say the same dumb thing again. At worst, I won’t have to ask Sam from Bangladesh to help me set up my WiFi anymore. At best? Too soon to say.
Comment from dissent555
Time: January 28, 2025, 11:16 pm
yawn …
… sips tea …
……turns page of hardcover book ……
Comment from S. Weasel
Time: January 29, 2025, 11:03 am
The code is open source; if you sign up to use DeepSeek, the Chicoms even tell you they’re monitoring *everything* – including your typing patterns.
Comment from Jon
Time: January 29, 2025, 8:54 pm
Some Vegetable, it’s not that I don’t think it will eventually have some use, it’s that right now it’s being called the Best Thing Since Sliced Bread without any actual proof. If they could settle down and describe what it’s actually good at (if anything) I might be inclined to listen.
Write a comment
Beware: more than one link in a comment is apt to earn you a trip to the spam filter, where you will remain -- cold, frightened and alone -- until I remember to clean the trap. But, hey, without Akismet, we'd be up to our asses in...well, ass porn, mostly.<< carry me back to ol' virginny