@Strawman sounds like with a library of standard queries it might provide some competition for the likes of simplywallstreet.com. (This allows a small number of free reports every month, and I sometimes use it for a quick scan.)

Agreed @mikebrisy Especially when these models can also produce charts (some can, i believe. But not Claude at this stage)

@Strawman ive tried a couple of times to generate company reports, which the AI models can do, but are not at the level of insightfulness that i hoped for. i feel there is a skill involved in prompting the models to get the desired response (as you say) and that i have yet to learn by the looks of it.

Yeah, it's still early days. What's fascinating is how quickly they are getting better. We've had the zero to one moment with these models, now comes the refinement and optimisation, and maybe a few bigger breakthroughs along the way.

Having a well-versed, financially literate AI model that can act as your personal assistant analyst doesn't feel that far away. Even at the current state of the art it feels like it at least represents a highly flexible and powerful search tool.

For those that like to transfer data to spreadsheets, there's also a lot of value in being able to get the AI to convert graph data into a table.

Eg I uploaded this table for AVA's latest preso, and asked Claude.ai to give me the figures

Based on the image, Claude.ai gave me this:

Sales Orders Received ($m):

Half-Year | Access | Detect | Illuminate | Total

2H FY22  | 2.8  | 5.2  | -     | 8.0

1H FY23  | 2.0  | 9.0  | 3.0    | 14.0

2H FY23  | 2.5  | 12.0  | 4.5    | 19.0

1H FY24  | 2.8  | 12.5  | 5.5    | 20.8

Percentage Changes:

Access:

2H FY22 to 1H FY23: -28.57%

1H FY23 to 2H FY23: 25.00%

2H FY23 to 1H FY24: 12.00%

Detect:

2H FY22 to 1H FY23: 73.08%

1H FY23 to 2H FY23: 33.33%

2H FY23 to 1H FY24: 4.17%

Illuminate:

1H FY23 to 2H FY23: 50.00%

2H FY23 to 1H FY24: 22.22%

Total:

2H FY22 to 1H FY23: 75.00%

1H FY23 to 2H FY23: 35.71%

2H FY23 to 1H FY24: 9.47%

When i eye-ball the chart, the numbers seem a bit off, and I did have to ask it to refine these figures a couple of times (again, i probably needed to prompt better). But its a cool use case.

When AI can accurately forecast the future I’ll be signing up. But then everyone will have the same information, the same valuations, and businesses would be perfectly priced. There would be no unforeseeable gains to be made, unsustainable loss making businesses would shut down in advance saving us from any pain. We would be so relaxed about investing that we’d have to take up BASE jumping for an adrenaline hit! Nah, not for me! :)

@Strawman I've noticed something odd with the free Ai services recently.

About 9 months ago, I could give Bard prompts to do quite a bit of data scraping. An example would be generating long lists of EPR software deals announced in the NHS.

I'd do things like "Thanks, that's a great list,... can you generate another 20 deals in the same format" etc. etc.

So last year, I could quickly and easily generate huge amounts of data in this way. But more recently, I'm getting more and more responses where the Ai gives reasons why it can't do it, and then it just give references to some web links and an instruction to essentially do it myself.

What are the possible reasons:

1. Trying to get you to move to the premium Ai service?
2. Managing the computational cost?
3. Maybe I need to learn more effective ways of prompting?

Is anyone else finding this?

@Rick I've also been thinking about this, too.

As people use Ai more and more, won't it just drive more and more of the market to the "consensus view"? With a narrower spread. For example, spreads in broker views are often down to differences in discount rates, terminal values, or the sector multiple involved, and relatively little to do with this years and next year's EPS or fundamentals.

Won't that actually create an opportunity if your investing ideas rely on generating deep and potentially contrarian insights?

For example, my outperformers nearly always arise where I've taken a longer term view than the analysts, who use reversion to come kind of average valuation multiple within 3 to 5 years. But for growth companies, very little of the intrinsic value actually lies in the next 3-5 years, other than setting the launch point for the subsequent 10 years.

If that's the case, then I can't wait for the world to fully embrace Ai fuelled investment decision-making. Bring it on!

You make a good point @mikebrisy. I agree on time horizon investing. My biggest returns have been from long term views, and there are several stocks I’ve sold too early after running out of patience.

I agree that most brokers think short term, and many broker views are based purely on share price trends and historical trading ranges (I won’t mention any names here, but there a few on “the Call” who I switch off to). I think short traders (shorters) often take short term views on businesses also, eg. Polynovo. They are often only looking one or two results announcements ahead.

I think Ai can replicate first order thinking really well and do it a lot faster than I can. The problem/opportunity is I'm not lacking for first order thinking. The less charitable might suggest that if it weren't for first order thinking, I wouldn't do much thinking at all. What we 'want' (and by 'want' I mean DON'T want) is for Ai to do 2nd, 3rd...nth order thinking. If it ever gets to that point we could be in trouble.

By the way, how does Claude get an Ai named after him? You know what software marketing teams are not contemplating naming their hyper-smart, life-changing, potentially-humanity-destroying Ai? Noddy! It's not on anyone's list...

Spot on @Noddy74! Although I had to use my first order thinking to look up what 2nd, 3rd...nth order thinking actually was! :)

I found this nice little 3 minute read to help me with my nth order thinking. I think (ie. first level) it was very interesting, and it’s what we do as investors to find an edge. https://fs.blog/second-order-thinking/#:~:text=First%2Dorder%20thinking%20is%20fast,let's%20eat%20a%20chocolate%20bar.

Noddy.ai is just around the corner I'm sure ;)

But i think these are all good points. I definitely think any informational edge is already long gone (excluding illegal insider info). So it's still a question of how we, the humans, investigate and interpret the data -- even if we're using more advanced tools to slice and dice it.

I mean, i can already do long division with a pen and paper, but i'll use a calculator thanks. So does everyone else, of course, and we all get value from it, but doesn't mean we'll all reach the same conclusions whenever their are subjective variables at play.

I do think compute costs and upgrade incentives are at play @mikebrisy. The front end models are probably just loss leaders for many of these companies.

LOL @Noddy74 , when I first glanced at @Strawman 's post yesterday, I thought he was actually referring to a conversation with Claude W. I hadn't heard of claude.ai before his post.

I wonder if you could ask claude.ai to write an investor note on $PME "in the manner of A Rich Life". Who know, maybe that's what Claude W is already doing.

My conclusion,... too many Claudes.