The study found that company implementations were failing while people heavily used their own choice LLMs.
> Over 80 percent of organizations have explored or piloted them, and nearly 40 percent report deployment. But these tools primarily enhance individual productivity, not P&L performance.
How on earth does "enhancing individual productivity" not improve P&L performance?
If the individual is more productive, they can work fewer hours. This is especially true in a remote-work environment where you are not in an office setting.
I have never seen people working fewer hours due to technological improvement, productivity just goes up from them working more efficiently for 8 hours.
And also, just because an employee is producing more code or more reports or more trades or more summaries per day, that doesn't guarantee the company will make more money. If you enabled your market analysts to make twice as many trades per day, you will likely lose money if they actually did that. Because they were already finding the good trades, and now they are also making the not-so-good trades and maybe even the unlikely trades.
And also, in the longer term not being shown in existing studies, you are in an arms race with your competitors. Like advertising, you will be spending a lot of that money in order to counter the spending of your competitors, and you both lose the game of prisoners dilemma, the only winner being the arms dealer selling to both sides.
> We’ve managed to convince the biggest tech companies to invest trillions into tech that doesn’t replace SWEs, rather, a new dev tool at our disposal to make some aspects of our job easier: search and tedious coding tasks, saving us time.
Another day, another substa...
Lots of discussion just a few weeks ago
https://news.ycombinator.com/item?id=44941118
slyzmud•1h ago
https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Bus...