[DanielleMolloy]

This blog gets passed around a lot recently. While I do draw value from the thorough observations of developments, the amount of text the author spends on shallow negativity can feel like the same waste of time as the overhyping PR machine he is reacting to.

There is without doubt something novel in the successes of convnets for sensory perception, deep Q-learning for decades-old and new game problems, artificial curiosity, recent machine translation, generative models and their various applications. Recent models also found their way into for-profit companies. It’s legit to be fascinated by this, and I’d rather stand on the side that doesn’t remain in their cave.

AI research may have picked all current low-hanging fruits or hit a wall either soon or in ten years, nobody can know yet, so there is no reason to run around predicting the future painted in only positive or negative light.

[bitL]

We need contrarian voices for both spotting any issues we might have overlooked, and assuring ourselves we know better.

It's still better than what I can read from "LinkedIn influencers" in my feed like "Logistic regression is still the best" or "Self-driving cars will never work because of long tail"...

[JesseMReeves]

It’s a good read but the negativity makes it appear irrational. It would be better if he left the ranting away and focused on the realistic recap without the PR hype.

Note that AI has a history of being stalled by overly pessimist evaluations (Minsky / Papert on the perceptron, Lighthill report).

[dreamcompiler]

AI also has a history of being stalled by believing its own hype (the AI winter of the 1990s). Too much money can kill AI as fast as too little.

[bitL]

Seems to be happening to VR right now...