Great framework @mikebrisy (which I'm totally going to steal!), and I reckon you're right with EOL example @Shapeshifter

@mikebrisy I had similar thoughts to you reading the email this morning, I was worried @Strawmanwas building up to a hiatus announcement. Warren Buffet has been saying the same things for the best part of a century, the beauty is that every year he finds a new audience, and for the repeat listeners, different parts resonate and different ideas click as their own understanding grows.

On reflection 2, my personal thesis on AI is it will struggle because something (or someone) who is correct 99% of the time is perversely more untrustworthy than something that's right 50/50 or even 75% of the time. If you know it may be wrong, you double check. If you're used to it being infallible eventually you'll take something incorrect into a high stakes decision and get burned. AI is also training itself to appear correct, so it misses a lot of the red flags were used to when dealing with people. I think it falls into the category of "talks the talk but can't walk the walk" until it can solve that problem (or people gain experience and get better at driving it).

Mike to be truthful I was shocked when I moved from professional to retail investing how much people relied on google, for “research “, something I had never done. I suppose I had the luxury of research that could be attributed to someone that had their reputation tied to what they wrote. Otherwise I see ent to source documents, everything else was just opinion. Even how I feel uneasy relying on people I don’t know for facts. Maybe that reluctance has reasons lol

Well, I found the Strawman article to be compelling reading - almost like a detective novel where the murderer is uncovered on the very last page.

Take heart Strawman, your issue isn't rare. I wrote a monthly subscription business newsletter (Positive Business) for 19 years and many was the time a subscriber would say 'I know that' to which my standard response was "I know you know, that's not the issue, it's what have you done with that knowledge, because knowledge without action is the same as not knowing at all" - 99% of the time it was game, set and match to me.

On the basis of common sense is not all that common, we all need to be constantly reminded of the basics and specifically, are we applying them.

I always find your weekly newsletters very handy and of a quality rarely found, so I think you are being a tad hard on yourself.

Great points by @Lewis and @Solvetheriddle. I always treat the output from AI searches with suspicion in proportion to the importance of the search or the type of decision it support. I find it pretty reliable in summarising or extracting insights from a known dataset, like a set of annual reports or a suite of provided documents. But when I task it with work where it gets to chose its sources, I always personally check any sources which are critical to the key output. I do this in the way you would carefully check the work of a newly hired BA, who hadn’t proven themselves to you. Of course, in your prompts you can direct AI as to the source quality, and I find that a helpful thing to do sometimes.

Like you @Solvetheriddle, I’ve had to make the transition from well resourced organisations where you could access high quality information resources to being a sole operator. And yes, Mr Google has been a key resource for many years now. But with effort you can still often find quality information resources if you put the effort in. It’s usually a case of thinking carefully about what you really need and triangulating questionable info from multiple independent sources, and hopefully finding quality information resources if you look deeply enough.

The AI anecdote I shared showed how I wasn’t prepared to pass on a low quality source, or an AI summary on its own, without doing my research due diligence. But it was a "near miss" because initially I had honestly been duped, and in my initial verbal communication, unwittingly became part of the chain of misinformation.

It was a wake-up call because I take real money decisions based on the strength of my analysis. And as the old saying applies … GIGO.

As @Lewis rightly observes, generative AI is skilled at dressing up garbage and making it look like a quality product. So I will consciously redouble my efforts to take some of the capacity AI has given me to make sure I do a thorough QA/QC whenever it matters.

@mikebrisy, interesting take on using AI and I've always liked your BA analogy, this seems like a good way to frame it to me. It feels like the BA has become more eager to please over time so you need to really ask for the risks / negatives when looking broadly at a business.

The sources are of course critical so using constraints seems like a logical if not foolproof approach to controlling for garbage sources.

I’ve similarly been duped by sources that when I looked through to them found it was AI generated slop – in my case a page giving ASX research that I’d never heard of but looked good. Lucky for me they had mixed the attributes of 2 companies into one and I knew enough about the business being discussed to know some of the info was way off and must have been about a completely different business.

‘Closed’ tools like Google’s NotebookLM are good for this as it only uses sources (PDF, MP3, URL’s, etc) that you give it. ChatGPT has custom GPTs for this, Perplexity has ‘spaces’ or something like that, Gemini has Gems, etc.

It’s horses for courses – sometimes you want AI to go fishing with a net and you need to tune your BS meter accordingly. Other times you want spear fishing / precision in which case you might be treating AI like more of a search tool, or used closed models with only info you give it.

Another trick I’ve just started using is in ‘Open’ models like Gemini, ChatGPT or Perplexity, etc, prompt it with something like "Using only ASX announcements....and the attached..." that way you are constraining it to info from sources that should be reliable / controllable.

I also include at the end of each prompt for this sort of thing “Show references to source material wherever possible.” and “Take your time, I’m in no hurry.”

Not sure if the last one helps but worth a try in case there’s a trade off between getting it done quickly and getting it done accurately like with actual BA’s.

This is a great topic.

While we are calling out the errors that can and do occur when relying on AI research… how many folks (not saying anyone in this site, I’m referring to the investment community at large) have followed the advice of ‘experts’ and lost a material amount of money? Human advice is not always great, and therefore, DYOR on the source materials and everything else outlined in earlier posts is still very much needed. As Strawman has highlighted a few times, you can borrow the idea, but you can’t borrow the conviction!

On a somewhat related topic, I also can’t help but think about some of the commentary regarding self-driving cars and the subsequent discussions when they’re involved in an accident. I wonder if the pass mark is zero accidents, about the same as humans, or just fewer than humans? My initial thought is that if they are as good as humans, or even better, then that’s a pass mark. Zero accidents (aka mistakes) is not realistic or even need for the roads to be safer than they are now.

All of that is the long way of saying: There will always be a need to look under the covers, regardless of where the advice is coming from. After all, even Uncle Warren has been wrong a few times, so let's be careful not hold AI to a higher standard than we hold the humans...nor should we just assume it is always correct, not for now anyhow lol :)

My initial response to this is of fear and anxiety if a heavy reliance on these LLMs ensue, which day inevitably will because there's less friction to getting work done. There is severe risk that the US CIA and government will capture this opportunity as a means to coordinate masses create confusion and drive populace thought in order to keep themselves at the top of the pyramid of control as has been the case with the underlying monetary systems and the way they are coerced and manipulated globally.

That was an interesting listen @SayWhatAgain

I think Yanis made a few factual errors (eg tether buys short dated, not long dated bonds, and I can't see how Japan would ditch US treasuries -- which pay a yield -- with USDT which is a zero yielding abstraction on the actual currency), but the thesis that Trump is deliberately trying to weaken the dollar is interesting.

Honestly, I get the strategic rationale of your goal is to reshore productive capacity. But it's just going to be a very long and expensive endeavour. Certainly could have been executed better, so far.

Certainly was @Strawman. I don't think he believes the Trumpian strategy will succeed, but he is right in warning not to underestimate what his regime is doing. His parallel to the Nixon shock was fitting, and his commentary that it is not the man but the team behind that we should not underestimate makes sense to me. I agree that his suggestion that Japan could be bullied into carrying out the dollar devaluation plan, potentially via crypto, sounds far-fetched...but it is provocative. His comments on China, I think, are spot on though...they do have a rapidly emerging digital infrastructure and if they get pushed too far, they will 'push the button' and still keep the trillions of US dollars they hold....and that is when we will see a change in the world order!

This is another (more pessimistic) take on Trump's plans by Satyajit Das (Ex-financier warns economic downturn comparable to Great Depression could be beginning abc.net.au). He also suggests they have a plan, but one that is somewhat more erratic and destructive. Both Y and S make the same overall point though, we are in the midst of significant, maybe structural realignments. What does it mean for us? he suggests that the effect of all this will be secondary, mostly because of our economic reliance on China. He suggests that the time will come when we'll have to choose. He quotes Kissinger and says that it is dangerous to be the US's enemy, but being their 'friend' is fatal. That may be true, but I don't know what the USA have done for us....and in a changing world order we should probably align with those in our region (?)....to be honest, personally I am not sure.....What concerns me, though, is how little strategic conversation there is on this from either side of politics. Our major parties seem adrift — uninspired, and uninspiring, more interested in grabbing headlines than showing much in the way of vision, let alone strategy....humphhhh!