Hacker News Reader

You might well be 100% right. I'd like to see the Japanese succeed, actually.

However, it was simply not possible for anyone to succeed at AI in the early 80's. It took some Nobel-prize-winning software, a change of approach, and a massive increase in compute power to finally break through.

View on HN

The only thing we needed was the compute. Everything else had been discovered by the 80s.

I think that we also needed a much larger training corpus than was available in the 1980s. Back then the largest textual data sets were orders of magnitude smaller.

Using neural nets. The approach taken towards AI in the 80’s was much less compute intensive.

Neural nets with hidden layers is literally older than the Apollo 11 landing. That doesn't make sense but it is.

Read TFA. The Fifth Generation Project involved expert systems & prolog, not neural nets.

Reply on news.ycombinator.com