An autobiography

I like to think I cultivate an air of aggressive unlikability, which may or may not be the way I deal with being an introvert. I wake up at 4:30AM most days — not because I have to, but for the…

Smartphone

独家优惠奖金 100% 高达 1 BTC + 180 免费旋转




Theory falling behind practice in Cryptocurrency and AI

As I am learning cryptocurrency and AI specifically deep learning, I found out that there is similarity in both of them with respect to recent progress in the field. In both fields, theory is following the practice i.e. something worked in practice better than what was expected from theory. This is not to discount the theory as the progress is based on years of theory backing them but practical approach worked before theory predicting it.

Bitcoin

In case of Bitcoin it is quite obvious as Bitcoin almost came out of nowhere without years of research progressing towards it. Cryptographic primitives and P2P protocols, I am sure were common knowledge among the practitioners of relevant fields but nobody thought to put them in the way Satoshi did.

> “When bitcoin first came out, I was on the cryptography mailing list. When it happened, I sort of laughed. Because I had already proven that decentralized consensus was impossible.”

Apart from this, Bitcoin solved (or is solving) many other problems like market dynamics, economic incentives. Some of these problems may have been solved accidently but it is matter of study now that it seems to be succeeding.

Deep learning

Even in case of deep learning i.e. supervised learning with multilayered nueral networks, practical applications proved that it is best approach before theory could give its verdict. Theoritically deep neural networks should overfit the problem but they seem to be doing surprisingly well. Of course, huge data is helping avoid that but deep neural networks would help solve the issue wasn’t immediately obvious.

Only after deep learning took off that theoritical explanations are coming up to exaplain its success.

Finally, theoritical work is very important for progress as Bitcoin and deep learning is based on decades of theoritical work but it might not be the only path to getting something working.

Add a comment

Related posts:

New 2018 IDE Video Released

Learn about all of the ways the team at the MIT Initiative on the Digital Economy (IDE) is helping organizations and people thrive in an era of profound digital transformation.

car insurance 2 days

I recommend you to visit this internet site where one can get quotes from different companies: http://insurancefastfinder.xyz . By continuing we assume car policy Below we’ve kinds of excess…

Agonizante

Todas sus pesadillas, incontables y atemorizantes pesadillas tomaban su cabeza a esa hora de la noche en un rincón de su habitación, luces apagadas y una música ambientada en algún tipo de funeral…