Hi @topowl

I've tried paid subs for both chatGPT 4o and Claude (by anthropic). Both pretty decent, and a lot better than the free models.

That being said, they do hallucinate a lot so you need to verify what they spit out. But a good follow up prompt can help with this (eg point out what page you found this information on).

It's been a game changer for my workflow. They won't think for you, but think of them as an eager intern that can help you with basic tasks.

@topowl I use Perplexity.ai, which allows you to run 5 free "Pro" searches per day and provides excellent visibility of sources.

I also use free version of ChatGPT, Claude and Gemini (Google, fka Bard) but find myself using the last of the three less often - I think it is too lazy and they want you to use the Pro version.

Agree with @Strawman in that it is an absolute game changer in terms of my research productivity.

I'll eventually sign up for a Pro model, no doubt.

Thanks @Strawman @mikebrisy

Appreciated

I don’t suppose @Strawman and @mikebrisy you can give some example use cases of things you find yourself doing with them on a more regular basis? It’s like 1999 and leaning how to best use a search engine…. Soon there’ll be a LLM clone for ‘let me google that for you’ .

Sure, just thinking about the last week, here are some examples:

1. Download a company investor transcript (available via various sources that use S&P Capital IQ) into a PDF file, upload the PDF file to AI and ask the AI to do things like: summarise the attached transcipt, focusing on "X" and ignoring information relating to "Y" and "Z". I'll often do this to analyse a competitor to one of my holdings, where I want to focus on information relating to the specific area of direct competition.
2. Or, I might want to have the summary focus on key information in which case I will use other keyword prompts and ask the AI to use these prompts as headings.
3. Generate a list of competing product offerings in a market I am looking at
4. Ask AI to estimate market sizes and market shares of the market leaders (mixed success here, but sometimes uncovers good sources)
5. Ask AI to explain complex issues that I don't fully understand, using terms that would be understood by [insert prompt]

I did all these in the last week.

Sure @Trancer

I'm the same as@mikebrisy -- Summarizing, searching and at times explaining transcripts and documents is probably the main use case for me.

I also find it good for suggesting article ideas, and outlining a basic structure. Also handy to get them to proof read also. They're not good at writing them though, at least not yet.

I've had a crack at using them to analyze financial statements, and they do an ok job -- but again you need to verify everything yourself. But they can pick out things you might otherwise miss.

@Trancer a growing required skill is prompting the LLM with the right questions, and follow-up questions. i saw a job advertised for such, i suspect that it is a case of the more you use it the better you get at generating anything useful out of them.

They won't think for you, but think of them as an eager intern that can help you with basic tasks.

That's how I tend to see them too. You really have to lead them by the nose, but you can get a feel for what they are good at reasonably quickly. I prefer to use them to help me find things rather than do any sort of analysis. There is still far too much "often wrong, never in doubt".

@Strawman I've given up on using AI for analysing accounts for now. I prefer to upload manually via an image into a spreadsheet, as the errors are easier to spot and its faster than checking an AI output. Too many mistakes on character recognition when reading PDFs.

Also, reading graphs onto data tables is often got badly wrong as well.

I am not sure if Aussie companies are in scope, but from what I have seen, FinChat.io is very good for (US) company analysis with an in-built chatbot.