Comment on 95% of Companies See ‘Zero Return’ on $30 Billion Generative AI Spend, MIT Report Finds

<- View Parent
ilinamorato@lemmy.world ⁨16⁩ ⁨hours⁩ ago

Please let me know what major breakthrough has happened recently in the machine leaning field, since you’re such an expert. Throwing more GPUs at it? Throwing even more GPUs at it? About the best thing I can come up with is “using approximately the full text of the Internet as training data,” but that’s not a technical advancement, it’s a financial one.

Applying tensors to ML happened in 2001. Switching to GPUs for deep learning happened in 2004. RNNs/CNNs was 2010-ish. Seq2seq and GAN were in 2014. “Attention is All You Need” came out in 2017; that’s the absolute closest to a breakthrough that I can think of, but even that was just an architecture from 2014 with some comparatively minor tweaks.

No, the only major new breakthrough I can see over the past decade or so has been the influx of money.

source
Sort:hotnewtop