Has the market peaked?

economic

GreatApe wrote: »

Humans are just made of about 1.2kg of brain there is no reason to suppose by chance humans evolved to have the absolute peak of intelligence theoretically possible. We are limited by DNA and also the size of our skulls. It is just a question of when not if humans create intelligences greater than our own.

An artificial intelligence can be electronic and run at 5 GHz rather than the brains 100 Hz
The artificial intelligence can be housed in a warehouse bigger than the worlds biggest building with billions of chips if necessary.

Moore law has already jumped a few times from one medium to another. First there were relays which got better and better. Then we jumped to vacuum tubes which also got smaller and better. Then we went onto silicon transistors and that has been great for about 40 years and maybe has another ten years to run. Its quite likely we will move onto another medium for information processing if we come to a point the current medium is not growing. Semiconductor industry is $500 billion globally there is a lot of money and resources to do the R&D needed to keep improving

I think its primarily a software problem at this stage.
If AI needs a computer a billion times faster than we have now
Well if the software was ready a company or government could say let's just build a billion computers in a huge warehouse rather than waiting 30 years for a single computer to become a billion times faster.

So if we are limited by software, how do we know we are intelligent enough to design the software to create the AI that is more intelligent then us?

Perhaps we need to evolve still further by using more of our brain to be able to produce the software that gets us there?

The AI stuff out there is just pretend AI. It is all based on big data and algorithms, that's all it is. We could have done this a decade ago as the hardware capabilities existed back then and so did software.

Perhaps we need to also evolve existing programming languages to be able to do this. We have moved from basic, fortran, lisp decades ago to now java, C, python etc. Maybe we need more powerful programming languages built on existing ones available to be able to take us to the next level.

westernpromise

The acid test of whether AI is any good or not isn't the Turing test. It's whether a machine can create fine art.

A painting, music, a film, a novel. Something with newness. You could probably programme a computer to generate Impressionist paintings but what would be evidence of human like intelligence is if the computer looked at Impressionist paintings, shook its head and invented post-Impressionism.

How would a computer stop producing this:


and decide to start producing this?


A computer can be a camera but that doesn't make it a photographer.

GreatApe

economic wrote: »

So if we are limited by software, how do we know we are intelligent enough to design the software to create the AI that is more intelligent then us?

Human knowledge grows with time. It is also a massive prize perhaps worth in excess of $100 trillion so lots of people are going to spend lots of time thinking about it and doing R&D to crack it.

The AI stuff out there is just pretend AI. It is all based on big data and algorithms, that's all it is. We could have done this a decade ago as the hardware capabilities existed back then and so did software.

From my understanding Google with its Alpha Go team seem to have the beginnings of a general AI on its hand. Likewise a company called open AI seem to have something similar.

Yes it needs to train by doing and learning from its mistakes but how can it be otherwise.
It is said they need huge amounts of information and humans only a little bit of information to learn new things. But I think that is a mistake humans need a huge amount of information we just dont realize that seeing is huge amounts of data. If you show a person who has never seen a car a car for the first time they dont just have 1 picture of a car. If they look at it for 5 minutes they have 60,000 images of that car at different angles and distances etc go through their eyes to their brain. So what looks like small data...just show a child a car once and it knows what a car is...is more than just small data it is a lot of data perhaps in the 100 GB range

GreatApe

westernpromise wrote: »

The acid test of whether AI is any good or not isn't the Turing test. It's whether a machine can create fine art.

Already there

Google ran its AI networks backwards and it comes up with art like this

http://www.popularmechanics.com/technology/robots/a16089/google-neural-networks-daydream/

economic

GreatApe wrote: »

Human knowledge grows with time. It is also a massive prize perhaps worth in excess of $100 trillion so lots of people are going to spend lots of time thinking about it and doing R&D to crack it.



From my understanding Google with its Alpha Go team seem to have the beginnings of a general AI on its hand. Likewise a company called open AI seem to have something similar.

Yes it needs to train by doing and learning from its mistakes but how can it be otherwise.
It is said they need huge amounts of information and humans only a little bit of information to learn new things. But I think that is a mistake humans need a huge amount of information we just dont realize that seeing is huge amounts of data. If you show a person who has never seen a car a car for the first time they dont just have 1 picture of a car. If they look at it for 5 minutes they have 60,000 images of that car at different angles and distances etc go through their eyes to their brain. So what looks like small data...just show a child a car once and it knows what a car is...is more than just small data it is a lot of data perhaps in the 100 GB range

Long GOOGLE

Crashy_Time

ilovehouses wrote: »

I spent a long time thinking about that. In the end it wasn't that I thought I had a UK edge but a realisation that I'm not getting any younger and will be requiring the money as sterling so thought the increased UK exposure mitigated the currency risk.

Also, as I sold single UK shares I took the view that a UK tracker whilst still being exposed to the UK was at least more diversified. I really wanted a FTSE All-Share tracker but went for the FTSE 100 because it was quite a bit cheaper to hold in my pension account.

If I was 20 I'd be 100% VWRL.

Isn`t that still allocating money like you have an edge though, although the currency point is a good one, wouldn`t UK gov. bonds diversified over different maturities be better as you are gambling on global/US markets not pulling down UK markets?

economic

Crashy_Time wrote: »

Isn`t that still allocating money like you have an edge though, although the currency point is a good one, wouldn`t UK gov. bonds diversified over different maturities be better as you are gambling on global/US markets not pulling down UK markets?

You don't need to have an "edge" to allocate money. You do whatever allocation suits you. If you are young, its best to be overweight technology stocks as a hedge for your future (in case tech takes over your job). Whilst this seems like having an edge, it is not, you are merely allocating to suit your personal circumstances. If you are older, you would have more money in defensives, home currency and income based. This doesnt mean you have it all in this, its just that you have more of it then if you were 20 or 30.

economic

ilovehouses wrote: »

I suppose so yes. The choice was between more world tracker and FTSE tracker - I want to be >80% equities so more gilts weren't an option.

In the end I'd dramatically increased diversification so stopped trying to overthink it.

How old are you?

GreatApe

ilovehouses wrote: »

You don't need an edge but you're claiming one by saying technology stocks or dividend stocks are best. You've claimed an edge because the world allocates capital differently and you know better.

You acknowledge that young people can afford to take more risk but, rather than claim an edge, both age groups could invest in the same world tracker but the oldies could mitigate the risk by holding a 'zero' risk asset as well. There's no such thing as zero risk but we're talking about home currency gilts.

All companies are going to eventually become technology companies

westernpromise

GreatApe wrote: »

Already there

Google ran its AI networks backwards and it comes up with art like this

http://www.popularmechanics.com/technology/robots/a16089/google-neural-networks-daydream/

That's not art, however. It's not a conscious attempt to create art or even simply newness. What, for example, is the artist's view on how light should be represented, and how does this differ from other schools' views?

It's art in the same sense that this is sculpture: