ChatGPT

[specuvestor]

Anyone tried asking best stock to own in SG, Asia, U.S. etc or if Nvidia will reach $1000 next year kind of angle? 😉

[Wildreamz]

Not recommended because ChatGPT only has limited access to the web (through Bing API). Can try with Bard, also not recommended because they are inherently not fine-tuned to give financial advise. That said, a fine-tuned model will likely be released before long..

[weijian]

The ability to calculate discounted cash flows lost its edge once everyone could do them on a spreadsheet.

Cigar butts became no more once everyone could grab hold of financial statements easily.

Will ChatGPT's output on "what to buy/sell/do" in stock markets be competitive enough to make BIG money, if everyone can sign up for it (for free) or 20bucks per month for the advanced version? Maybe regenerative AI is bigger than an adaptive complex ecosystem like the stock market? 

[EnSabahNur]

[quote pid="168437" dateline="1687145663"]
The ability to calculate discounted cash flows lost its edge once everyone could do them on a spreadsheet.

Cigar butts became no more once everyone could grab hold of financial statements easily.


[/quote]
I guess the way forward for us is either:

1) figure out how to use generative AI to do what we do, but better (which many of us are doing)

or

2) figure out what generative AI cannot do, which is an arms race as generative AI gets better and better

3) find the weaknesses of generative AI and bet against it

[specuvestor]

I think data is not king. I think the use of information (derived from raw data) is king

Equities are not like bonds which is very mathematical. Even for DCF you have to guesstimate the discount rate or WACC which entails opinions. Computers can do repetitive things very well but AI is supposed to intelligently decide on unusual environments, not data dump. In short human heuristics which is not perfect but helps us decide whether to catch a bus or not

In fact if ChatGPT is not predictive model but rather language model, then it will likely take the consensus view which is probably not VB view. But nonetheless Im curious

[weijian]

Machine learning was once about using your own data to create predictive models. So a model to using your own data to predict turbine failures for P&W engines could not be used for Rolls Royce's.

Interesting to see how this turns out as a media giant decides to "test and stretch" the potential synergies between "data" and "use of information".

The New York Times sues OpenAI, Microsoft over copyright infringement

THE New York Times sued ChatGPT-maker OpenAI and Microsoft in a US court on Wednesday (Dec 27), alleging that the companies’ powerful AI models used millions of articles for training without permission.

Through their AI chatbots, the companies “seek to free-ride on the newspaper’s massive investment in its journalism by using it to build substitutive products without permission or payment,” the lawsuit said.

With the suit, The New York Times chose a more confrontational approach to the sudden rise of AI chatbots, in contrast to other media groups such as Germany’s Axel Springer or the Associated Press that have entered content deals with OpenAI.

https://www.businesstimes.com.sg/interna...fringement

[LionFlyer]

Machine learning was once about using your own data to create predictive models. So a model to using your own data to predict turbine failures for P&W engines could not be used for Rolls Royce's.

The issue stems partly from whether the output of generative AI is considered copyrightable. Under current US laws, that is no. So using your example, RR will not be at fault or violate any laws if they used the output of PW's gen AI.

If you use DALL-E to generate some logo or marketing materials, you can't copyright it, unlike that you typically do when you go to a marketing/advertiser and sign a contract where the foreground IP belongs to you.

Companies naturally will not allow their own data to train such gen AI models as there isn't anyway to monetize that downstream output.

edit: as a side note, it is ironic that Chinese courts are 180 degrees on this and recognizes the IP arising from gen AI.

[weijian]

hi LionFlyer,

Apologize for the confusion. I was actually referring to PW's engine data for ML will only be applicable for predicting failure for PW engines. But the same data or model from PW is totally useless when it comes to predicting RR engines' failures.

[Porkbelly]

Happy New Year to all Valuebuddies..!!

Thank you for all the sharing, the opinions, the perspectives.

May 2024 bring joy & happiness to all and at home..!!