Gen AI: too much spend, too little benefit?
goldmansachs.com
goldmansachs.com
Isn't that equivalent to saying on heads I buy, on tails I sell? What makes that incredibly good?
For example, invest $1000 in two companies, and expect a 100x gain and a $1000 loss over the next 5 years.
(I say "out of the gate" because you can change your bets as situations and positions evolve, and I'm sure MS do.)
I don't think Gen AI will be totally bust, but it won't be as promised (anytime soon). Just like in a software project, the last 10% is another 90%.
Reading a two column PDF on even a high end iPhone is not a pleasant experience.
(And no, I'm not going to break out a laptop to read something like this.)
Love your blog!
2) Much like a research paper, reports like this benefit from the standard page size. You will fit relevant information on the same page for the benefit of the narrative/reader. This is one of those areas where I believe its hard to make it work on a mobile device.
3) This is just the standard for reports.
I reckon a new generation moving into influential decision-making roles will probably sort this out over the next decade.
Similar to a powerpoint presentation, pdf reports are not going anywhere.
There's both too little real estate to present information and too few control inputs to manipulate it.
Keep in mind it came out about 1.5 years ago. That is insane market capture in a wide diversity of ways already.
LLMs are clearly more useful to the average user than blockchain. But there’s still an amount of over-eagerness.
It also excels in another kind of spam which is to generate articles to drive traffic from SEO.
But one can also argue that actively seeking and choosing targets is a generative task, given a proper encoding of a battlefield situation and current mission.
[0] https://en.wikipedia.org/wiki/Attacks_on_civilians_in_the_Ru... [1] https://www.ohchr.org/en/documents/country-reports/attack-fu...
— Certainly! bombs own civilian hospital
— No, the hell are you doing…
— You are right. I apologise for the oversight.
This system has no imbued morals or ethics, and the more we try to imbue some to it the more it fights back (again, the ad nauseam "less regulation" repeated over here). We need a new system that evolves this bullshit into something more humane.
It's not like the USA with its hypercapitalism, and absurdly immense wealth has achieved the greatest utopia on Earth so we kinda know capitalism is not the final answer.
Green energy already sees nearly $2 trillion in yearly capex: https://www.iea.org/reports/world-energy-investment-2023/ove...
Total previous green investments are $8-10 trillion, while fossil fuel consumption's not gone down (fossil fuel investment is falling, though.) That Stanford study undercounts and presumes a massive increase in storage.
If really dig in, our current renewable layout focuses on the low hanging fruit, but storing solar energy till the night time or a different part of the year is infeasible (look into "duck curves") and currently done by burning natural gas. Current storage production is up a full magnitude, but we need multiple further magnitudes more to do a full transition. The increase in transmission lines alone is more than all known copper (mined and unmined).
1) Automating boring reading and writing tasks. Think marketing copy, recommendation letters, summarizing material, writing proposals, etc. LLMs are pretty good at this stuff but these are not many people's core job responsibilities (though they may take up a lot of their time). Consider it a productivity booster for the most part. Some entry level jobs will be eliminated, and this may create problems down the road as the pipeline of employees to oversee LLMs erodes.
2) Code writing tools a la Copilot for certain "boilerplate" code in commonly used languages. I think the impact is similar to (1) where entry level jobs erode and this may impact employee pipelines.
The core problem (as I see it) is that LLMs don't produce outputs good enough to be used without human oversight except on a small subset of tasks. So you end up needing humans (maybe fewer of them) to check the LLM output is headed in the right direction before you let it out into the world.
Consider voice interface LLMs for customer service. When will they get good enough to do the job with real money on the line? If your airline help desk keeps giving away free flights or on the flip side infuriating passengers by refusing allowed changes, can you really use it in production? My sense is they aren't good enough to replace the usual phone tree just yet.
When accuracy doesn't matter that much, LLMs will really shine because then they can be used without a human in the loop. Think some marketing/advertising and especially, especially propaganda.
I think the existing killer apps don't yet have enough money/savings in them to justify the spend. If generative AI technologies can get good enough on the accuracy front to remove humans from the loop in more contexts, we will be talking about much more dramatic value.
We will in fact be talking about the most valuable thing ever
If they can't handle new ideas humans will always be much more useful and these systems are good for references and human learning are not good for creating something new and of value. I've noticed for text LLMs are quite weirdly repetitive and have an empty style that requires a lot of editing to get it into a shape humans would craft.
People will say the improvements are coming but I think most of them have come from more data which is running out. I think one of the most profound things about real intelligence is being able to define and update concepts within your own mind… how to add new information to LLMs in realtime and have that reflected across the board seems intractable given the training and refinement these things would sit upon. There is no clear unit of information about a concept that links to all the other ideas. LLMs seem quite limited by this.
The brain is so much more complex than these algorithms too and so much more flexible, I don't see how a very good encyclopaedia with some fuzzy AI concept extraction capability is in any way the same as the human brain being able to apply and adapt concepts from all around art, science, literature and the human experience.
If you're using ChatGPT, try adding "use Python" and see what happens.
Tool usage like Code Interpreter dramatically increases the capabilities of existing models.
NVDA is a decent proxy for OpenAI, their market cap would evaporate if big tech stopped buying.
You could make the argument that rate hikes caused both, but NVDA was looking iffy in between crypto mining and ChatGPT.
I did not say I would short NVDA personally, mind you.