# DL/ML impact through financial system
⏎
## Signal and Noise
⏎
- positives
⏎
markets as the engine of current capitalist-based society
⏎
- negatives?
⏎
no long-term perspective
⏎
- compare with "Noise" by Fischer Black
⏎
- [my thoughts on Twitter ahead of the Aaron's article](https://x.com/MTurlakov/status/1747394938868785337?s=20)
⏎
<blockquote class="twitter-tweet"><p lang="en" dir="ltr">6/6<br>Market power to<br>a) non-linear couplings - trading, observing, reinforcing <br>b) data owners<br>c) experimental hard information not yet digitised<br><br>Many challenges<br>a) autocrats in em<br>b) slow big bureaucracy in dm<br>c) people creating a lot of obscurity intentionally</p>&mdash; Mihail Turlakov (@MTurlakov) <a href="https://twitter.com/MTurlakov/status/1747394938868785337?ref_src=twsrc%5Etfw">January 16, 2024</a></blockquote> <script async src="https://platform.twitter.com/widgets.js" charset="utf-8"></script>
⏎
## 2024 - motivated by A. Brown "ChatGPT may rival Flash Boys in Transforming Markets"
⏎
1. The basic problem is that financial prices are nearly all noise, they are very close to random walks.
⏎
2. Traditional AI is more successful when signals are stronger relative to noise
⏎
3a. High-frequency trading, introduced in the late 1990s, ....
⏎
...  in Flash Boys. It led to zero-commission
brokerages and zero-fee index funds — eliminating the revenues that brokers and asset
managers had relied upon
⏎
3b. ...  in 1973 created the modern global derivatives economy, which vastly expanded leverage
⏎
3c. ... we’ve seen dramatic effects from credit default swaps, collateralized debt obligations and exchange-traded funds
⏎
4. ... is not ‘Eureka’ but ‘That's funny...’” 
⏎
consilience of E. Wilson 
⏎
**The Google paper suggested that AI should spend less effort figuring out which funny facts were important, and more time correlating all of them**
⏎
5. But there is **little useful theory or reliable quantitative generalizations about how stocks, bonds, real estate and other asset classes** should be priced relative to each other.
⏎
A plausible near-future story is LLM-flavored trading models will build large cross-assetclass portfolios similar to what global macro hedge funds do, but with more leverage, more positions, more active trading and no human to explain the thesis. There might be an explainer module added that will give plausible-sounding theses, but there’s little reason to believe these explanations will have any relation to the reason for the positions
⏎
6. I expect at least as much disruption as we got from HFT, and perhaps as much as we got from public trading of financial futures and options
⏎
And if I’m wrong, if LLMs and attention modules fail to gain much trading traction, there are plenty of new ideas in the AI pipeline to take their place.
⏎
⏎
⏎
⏎
⏎
## 2023 - Forget AI. The Real Risk Is the Dumbing Down of Markets
⏎
Some people dream of replacing smart humans with even smarter artificial intelligence that lacks human
behavioral biases and sometimes perverse incentives. That may or may not be a good idea. But replacing smart
humans with dumb computers clearly has more downside than upside for markets
⏎
- 2023 - AI Won’t Beat the Market Any Better Than Wall Street - Analysis by Nir Kaissar
⏎
“The ones using AI first may be able to uncover anomalies and exploit them. But once discovered, those
anomalies will disappear as others replicate the strategy,” Swedroe told me
⏎
If anything, AI is more likely to burn investors than benefit them
⏎
...  index funds increasingly mimic traditional styles of active management, such
as value, growth, quality and momentum. Still, if AI takes over for active managers, maybe indexers can finally
track broad markets in peace.     
⏎
# Parents
⏎
* The future of deep learning⏎