Back

@asymco this is going to be a multi year investment. Revenue is going to be increasing every year due to the AI being more mature and the initial cost will be paying it back.

@iamwaseem @asymco or maybe it won’t. Maybe it will be a value destroying boondoggle like self-driving cars, cryptocurrencies or the metaverse. I can see value in AI for some business productivity use cases, but there are absolutely no guarantees that this bet by the AI industry will pay off.

@karlstanley @asymco you have valid concerns. But AI is different if compared with crypto and Metaverse.
Forgot everything else, it is helping safety related products massively. It is making the world safe for everyone.

@iamwaseem @asymco how? How does a GPT or RAG or diffusion model make the world safe for everyone? The AI tools that summarise documents can’t do basic math. GPTs make up fake answers - there is no known way to fix this. Combine this with the issues around copyright theft of training data and the sheer energy cost of training models with trillions of parameters and it is by no means certain that the AI future Sam Altman is pushing will ever happen.

@karlstanley @asymco I didn't talk specifically about GPT or diffusion models. There are lots of AI models that are making world more safe. AI is a technology which can be used in good and bad ways, as any other technology.
BTW diffusion models are being used to generate reliable data which can be used to train models for safety related applications.

@unlambda @asymco The source being Nvidia‘s financial report that more or less states ~50bln revenue for 2023, implying that all of this came from the AI big shots. Doesn’t sound very realistic, honestly.

@fpbhb @asymco Here's an actual link to that source: https://investor.nvidia.com/news/press-release-details/2024/NVIDIA-Announces-Financial-Results-for-Fourth-Quarter-and-Fiscal-2024/

They do say "Full-year revenue rose 217% to a record $47.5 billion" for their data center revenue, and data center revenue is almost entirely AI, so the $50 billion might be rounding up a bit but is reasonable.

It's the $3 billion in revenue for the AI industry itself that I was more interested in sources for; that seems a lot harder to count.

@unlambda @fpbhb @asymco thanks for digging this up, Brian. I am also interested in a source, because this claim is confirming my priors a bit too much to be comfortable spreading it widely

@glyph @fpbhb @asymco That's exactly the reason I asked.

OP replied to me with this reference; unfortunately the WSJ is paywalled so I couldn't actually check the reference: https://mastodon.social/@asymco/112552691066961933

Most likely this is the article: https://www.wsj.com/tech/ai/the-ai-revolution-is-already-losing-steam-a93478b1

@kellogh @asymco @ivan3bx I don’t remember the book making that claim - what I remember is that incumbent firms (rationally) cater to their existing customer base and that leaves them open to disruption from below where competitors build inferior, cheaper products catering to the less profitable consumers ignored by existing players.

@karlstanley @kellogh @ivan3bx Indeed it makes no such claim. The pursuit of the more than good enough is the dilemma.

@asymco @karlstanley @ivan3bx chapter 6, under the heading “Leadership in Disruptive Technologies Creates Enormous Value”. in my version its pages 124-128

@kellogh @asymco @ivan3bx Awesome, thanks for the precise reference Tim - just checked and it's page 143 in my 2003 edition. Looks like I need to go refresh my memory!