Related Posts
Coach
Conversation Starter
How are hours at Simpson Thacher LA?
Thoughts on Trinity Partners?
Thoughts on Mindshare Chicago?
Have you heard? @Cleveland Clinic is opening a new hospital! Located in Mentor, Ohio, this BRAND-NEW state-of-the-art hospital will offer world-class care to Lake County residents.
We are looking for Nurses to join our team and be a part of Mentor Hospital's grand opening in 2023. If you are interested in learning more, please contact me. Schymak@ccf.org
Mentor Hospital | Cleveland Clinic
I have tons of mushrooms, what should I cook?
Additional Posts in Artificial Intelligence
New to Fishbowl?
Download the Fishbowl app to
unlock all discussions on Fishbowl.
unlock all discussions on Fishbowl.
Cheap, good AI means
1. The hundreds of billions big tech is investing in data centers and training AI might be bad investments because it can be done much cheaper.
2. Even with crippled Nvidia chips and limited access/skill for the most advanced chips, China is still able to produce competitive AI, negating efforts to limit its access to advanced AI by limiting access to the most advanced chips and chips fab tech.
3. Cheap AI means different products become viable, sooner. Right now, most AI is loss-leading--really expensive to build and deploy, with unclear paths to profits except for the people who host it or add it to existing b2b technologies or in very specific, narrow use cases (e.g., improving as targeting, enhancing trading algos). If you can make it cheaper, you can build and deploy it more, faster, and in different places. Suddenly, replacing people and tasks becomes easier, on-device AI becomes easier, and adding AI to services is cost-effective and profitable.
Caveat that it seems more and more like they lied, and actually, it was not very cheap to make at all, so 1 is probably wrong, and 3 is not necessarily going to show up instantly. 2 remains a problem, although a misguided one that we are wasting our time on, clearly.
This is a great point, thanks!
Alright, imagine you have a lemonade stand. To make the best lemonade, you need a really big and expensive machine that squeezes lemons perfectly. But only a few kids in the neighborhood (let’s say OpenAI and Anthropic) can afford this machine, so they sell lemonade, and everyone has to buy from them.
Now, one day, another kid (Deepseek) figures out how to make the same great lemonade with a much smaller, cheaper machine. This is exciting for two big reasons:
1. More Lemonade Stands – If making lemonade is now cheaper, more kids (or businesses) can afford to make and sell their own, instead of relying on just a few big lemonade sellers. This means more competition, more choices, and maybe even better lemonade for everyone.
2. Cheaper Lemonade – If you don’t need a super expensive machine, lemonade prices can drop. AI models cost a lot of money to train and run, and those costs are passed down to businesses and users. If Deepseek (or others who copy its method) can do the same job for less, then AI services might become much cheaper or even free for regular people.
So even if the lemonade (or AI model) tastes the same, the fact that it’s much cheaper to make could mean AI becomes more widely available, embedded into more products, and affordable for more people—maybe even free.
That’s why people outside OpenAI and Anthropic are also excited (or worried). It could lower the barrier for everyone to build AI instead of just a few big companies controlling it.
Wow so thorough, thanks!
For one thing, all the environmental stress that AI would exert because of enormous power requirements could be avoided if the work could be done with less input. So that's a help for society in general. And for the business world, it's obviously a huge thing. The investors who have been buying into the AI hype got a good shaking up over Deepseek. Now, the problem, of course, is how legit it all is. But assuming the breakthrough is real, it's a big deal.
Thanks for sharing!