I've been curious about others' thoughts on the over-reliance on AI for research aggregation and content generation. It's been cited of concerns of hallucination and biased content, but assuming the output is objective and in full, it presents a secondary problem: Skill Decay due to over-reliance. Any time you generate an argument or use it to aggregate research, you're removing a fragment of the decision making process by relying on it and for the younger audience, who haven't developed the judgement or skepticism to see through bias or discern false arguments, it can be detrimental to their learning and development.
I'm not refuting the immeasurable benefit of AI, but this industry looks to parallel the social media giants back in their infancy where no one would have imagined the negative affects they would have on adolescent mental health.
What's everyone's thoughts on this one?
Skill decay due to over-reliance on AI is a valid concern, especially in industries like finance and consulting where critical thinking, judgment, and nuanced decision-making are paramount. Based on the most insightful WSO discussions, here are some key points to consider:
Erosion of Fundamental Skills: Over-reliance on AI for tasks like research aggregation or content generation can lead to a decline in essential skills such as analytical thinking, problem-solving, and the ability to critically evaluate information. For younger professionals, this could hinder the development of judgment and skepticism, which are crucial for identifying bias or false arguments.
Parallels with Social Media: The comparison to the early days of social media is apt. Just as the long-term effects of social media on mental health were underestimated, the potential for AI to negatively impact professional development and cognitive skills might also be overlooked. The convenience AI offers could inadvertently discourage individuals from engaging deeply with the material or honing their expertise.
Bias and Hallucination Risks: Even if AI output becomes more objective and accurate, the risk of users blindly trusting its results remains. Without the ability to critically assess AI-generated content, professionals may propagate errors or biases, compounding the problem.
Balancing AI and Human Input: The key lies in using AI as a tool to enhance productivity while ensuring that individuals remain actively engaged in the decision-making process. For instance, AI can handle repetitive or time-consuming tasks, freeing up professionals to focus on higher-value activities that require human intuition and creativity.
Educational Implications: For younger audiences, it's crucial to integrate AI literacy into education and training programs. Teaching them how to effectively use AI while maintaining critical thinking skills can help mitigate the risk of skill decay.
Industry-Specific Concerns: In fields like finance, where judgment and skepticism are critical, over-reliance on AI could lead to significant errors. For example, analysts who rely solely on AI for data interpretation might miss subtle market signals or fail to question the assumptions underlying AI models.
The consensus from WSO threads seems to be that while AI offers immense benefits, professionals must remain vigilant about its limitations and actively work to preserve and develop their own skills. The goal should be to strike a balance where AI complements human expertise rather than replacing it.
Sources: Work at top AM -- AI will virtually kill of entry-level roles in AM over the next 5yrs, Will robots replace your consulting or financial career?, Thoughts on AI and the future of the job market?, https://www.wallstreetoasis.com/forum/hedge-fund/machine-learning-taking-over-hf-research-analyst-roles-in-near-future?customgpt=1, HYPSM ex-Quant Fund Intern Senior career advice and opinion on the future of finance and the world
well, I think you could have made the same arguments many times in history. You still need solid hard skills, but you don't manually do additions and multiplications instead of using excel. If you spend less time there, you move downstream, where everything is actually a) more interesting, b) more productive, c) actually harder intellectually.
I see it just way more interesting, but I totally get your feeling
This argument assumes good faith use of AI tools by the population, but I would argue that agentic AI models that off-loading more and more work to a solution creates a black-box organization where the work is getting done, but the cracks in the process become harder and harder to recognize or flag and this highler level strategic analysis will be poorly executed without technical understanding of the core functionality.
For instance, if I went to Claude and gave it a detailed idea of a python app I'd like to make, it could likely return a compile-able script. However, having never developed something like this myself, I'd be skimming or reviewing the code's high-level architecture and not fully capable of evaulating where the faults/gaps may exist.
I really like the equivocation to Excel because one thing everyone on this forum can personally vouch for is broken or archiac financial models that are skimmed or grazed that at best don't affect decision making and at worst, can grossly skew the valuation of a transaction due to some oversight (hidden cells not picked up, subtracting a negative, etc.). I imagine this type of occurence will exist in greater frequencies with greater reliance on agentic AI solutions.
The risk of black boxes is material, and I totally agree on this. Nevertheless, I think it's impossible not to end up there, and the full picture of industrial sectors with this tech deeply implemented I think it's still quite blurred.
Two main drivers:
A. Too much risk of losing out on this.
B. Too many incentives to move there.
Investors see the potential of big productivity leaps, and top management has to implement it.
If you look back, the first Oracle products were more a repository of bugs than databases, but then today would be impossible to live without. Or better, it's not that it's technically impossible not to use them, but your productivity is so much higher with them that you can't compete without.
Tech just needs time to deliver, but usually it does.
So I think the play here is more on how actively design these systems, more than their absolute adoption. Look at Harvey for legal firms. I think they are doing an incredible job.
Not a software engineer but a lot of large influencers on Youtube like Coding Jesus have claimed that with ChatGPT / AI / ways to cheat now in college has created many subpar graduates. Some of these influencers believe the horrible tech job market is partially due to the mismatched skillset of recent graduates.
Some jobs simply can be autmated now too. A mentor of mine worked on headcount planning at a company he's working at and it seems that with AI augmentation, they were able to cut a good chunk of their workforce to bring down costs significantly (he's at a PE Portco so profitability strategy is likely a lot more aggressive than most companies at the same size).
People like to cheat. Also during / right after the pandemic, from everywhere (ib/pe/re mostly) friends were telling me the quality of applicants was drastically reducing. We used a locked brower for exams, but at least 50% of the class was cheating using the most creative strategies. Now it's some ai tool. But then you can also use from the other side. Mercor is an example. You can do large-scale pre-screening using these tools, and manually review only the top candidates. Fun times
yeah I agree. Makes lazy thinkers. The brain has to be put under mental stress by exercising judgment and understanding what has to be done, which allows it to "unlock" new stamina to tackle harder problems. If you overrely on AI, especially at university, you may graduate with developing only 30% of the problem solving capacity that it is expected from a graduate. Because at the end of the day is not solving a math equation, is about solving a problem. If you have a fucking formula where you just need to input the numbers and you need AI, then what are you gonna do when you have a real life problem with many unknowns that requires a solution? Right, you'll remain mediocre.
This is something I was thinking weeks ago. That being from a generation that had to go through school/university/etc. without AI probably makes us more needed in a profession where human judgment will be needed. Younger people yes, they will leverage any possible tool they have and get the most precise answer possible, but then the question begs "so what?".
Because what can't be automated are unique perspectives that solve some problems or different scenarios, and not giving a number to a client that he can get it by itself with AI. Also, many solutions given by AI are extremely superficial to the point they are useless.
Let me give you maybe an exaggerated (but also not as much) example of what AI can't do. Let's assume that in some years there are some incredible software that basically could replace the entire RSSG PJT team. Their models will say the same as the software AI, RSSG PJT gets the same value on their model for ABC company as the AI model. But, then I come, i check the financial statements and I find a more creative way to interpret the GAAP/IFRS on 2/3 items on the balance sheet, which my accountant and tax lawyer approves. This increased my valuation perception (tangible, btw) to $200 by strictly interpreting some accounting limes in a new perspective that will also affect my tax savings or any other metric. Your AI software will confidently tell me that this company is worth X because it had run x1,000 valuation scenarios and projections, etc., to which will be DELOITTED to pay for it the X price, because I can get something worth X+$200 for just X.
Good luck to the AI hardos in "delegating" such skillset to a machine, because I can't wait to pick up your undervalued companies or explain to you that X assumption is fucking nuts - to which you would be unable to push back on it because you passed all your valuation class with ChagGPT and aren't even sure - and get your company for 15% cheaper. Even if you don't believe me, good luck spending extra costs for lawyers/accountants to negotiate with me. You just basically lost X value (minus) lawyer and accountant costs
very simplistic examples to get my point across, but in real life it can get very hairy
You're right, but I also assume you use excel. Does it make you less sharp not manually subtracting sg&as from gross profits? Or not having a floor of analysts calling all of day for the quotations instead of using bbg? It's provocative, but what I see it's not the substitution of a good mind, but freeing it from mundane parts of its work to put it at use where it counts. Then we have to set the border between the two, which is another story.
This is what I see: www.wallstreetoasis.com/forum/investment-banking/what-was-ib-like-in-th…
This one might be an interesting.
Issue is, modelling (and I use modelling interchangeably with “data gathering”) is generally a time sink. The pulling together of consolidated XLS files, adding informtion into the file from raw PDFs, neatly formatting the CFO’s rough work XLS et cetera. All time sinks.
Instead, if the “AI” (I usually refer to AI interchangeably with “processing / raw computing power”) can give me a first cut of a linked XLS that is neatly formatted, that itself for me is a HUGE win.
I would prefer to query the trends, WC spikes, unit economics, margin volatility, pass through costs, key customer risk, EBITDA add-backs and various other points to better understand the value and supply chain of assets instead of spending time pulling the raw data or formatting the XLS.
Once a dynamic file has been set-up, I am much happier reviewing it and tidying it (I don’t expect to rely on it for the final decision without checking and flexing it), as long it saves me time on the first cut.
But above is just the set up of the scene for the below.
Issue is, not a lot of juniors actually progress from the mechanical / robotic side of data gathering to “understanding” the data. And that, is precisely where the value add comes from. If the “first cut” that was being put together by analysts is now coming to me sooner and in a state that I can play with it, I may only need juniors for peripheral stuff.
It has been a major hiring concern now, as it is much harder to discern especially at the more junior levels. The way to fight that is more tests / less interviews (which can become a barrier for hiring). In other words, it really sucks to be recruiting right now because it is hard to tell if the person is great, or just good as copy pasting into ChatGPT. One massive reason why the job market is such at a standstill now (orange baboon would be the other reason).
Qui repellat eligendi corporis sint voluptatem. Molestias illo non quaerat ea qui. Nisi delectus sed consequatur dolor nulla. Reiciendis sunt quia at voluptas.
Pariatur molestias dolorem sunt rem ab beatae error dolorem. Corporis rerum soluta aperiam assumenda architecto dolorem sunt. Nulla eos et incidunt eligendi ut veniam. Est sunt error quis non fugiat unde quia. Autem temporibus cum accusantium magnam. Maxime eaque odit autem quas non consequatur dolor. Eveniet ipsam libero dolorem at.
Iure impedit error eveniet odit facere est molestias. Et eius velit voluptatem eius nesciunt inventore libero. Expedita sit aut asperiores voluptatem ut et. Praesentium omnis saepe qui ad non. Nemo explicabo repellendus molestias assumenda.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...