aisafetyceo

Opportunity mixed with difficulty.

I went down that path as conceptual learning for building a voice controlled website building system

- you can't count on making much money just reselling the software on larry pages computer because he is counting on making much money just reselling software on his computers ... https://teachablemachine.withgoogle.com/

Also.. more seriously, the latency to the google servers and back makes serious applications out of reach.

for ex. combine the concept with this https://comma.ai/shop and theres a geniune billion dollars waiting to be thought out

another ex. https://www.indus.ai but powered with many cheap https://amzn.to/2Xg4JkB

the path forward is to write a learning algorithm and recognize your work as one who creates virtual computers - screenless computing is a big deal, microsoft surface earbuds will have basic office integration, but imagine if you could your pipeline communicate the contents of pdfs,websites etc. without a screen around

theres the obvious consumer products but also imagine being able to work with a computer in new environments theres also the non obvious: imagine one smartphone camera pointed at a room, every person in that room can access their virtual computers using camera or voice You can sell kits that that transform rasberry pi's into useful robots

- 3d reconstruction is cleaner and more powerful then deep learning - stitching together screenshots of websites - learning as environment mirroring

its also useful to think of the practical economics of the distribution of the technology ... android phones can go around $70 - $1000 but they essentially run the same free software opensource software

so there is not much of dollars that any individual can extract from the distribution of "AI" software

What I would do is delve into the technical software reasons behind Googles acquisition of Fitbit.

Spoiler: How does fitbit allow developers to build with javascript and css without running a "browser"

The point here is Larry Page is eyeing the virtual software developer opportunity and so are you! in order to extract value from the vision recognition opportunity you must spend a lot of energy writing rules

it would be efficient if the end user could program their camera on the fly using natural language without internet connection on any device

this north star is the makers definition of an agi .. and as a independent business you can't afford to work on a subset of the problem

Unless your currently facing the issue and can't solve it using "true standards" than don't bother learning it.

I wouldn't use anybodies definition of a standard rather just try to get away with learning no "vertical software industries" which were pioneered to increase cloud vendor profits

I'm suggesting a low level stack of Linux EC2 with Nginx, Nodejs and no libraries

The vast majority of concerns that are addressed in books are solved by the invisible hand of the market.

a simple example of this would be "the tiny learning hell" a developer would place themselves through to implement SSL before certbot surfaced

the same is happening for security, caching, CDNS, Load balancing etc.

The component that I suspect is not in written circulation is what i think of as a single user model: across all systems the preferred way to implement a solution that scales is to reduce the complexity to a single value

For example, if you have database of users then you write the data to a single file with no real structure instead the structure is computed during a read request

Another example, if your writing a web app with a node server, reduce the server and webapp into one function on one page that runs through a generator contained in that one function that outputs the appropriate file for the server or the client

the value is that your code is reduced to the logic that produces real functionality and benefits naturally from the teams of engineers who work on improving the concerns and constitute the invisible hand

some insights that are noteworthy: - instead of using PM2/forever to keep your app online spawn a bash script to start your node process - Require.cache can cause issues / might make sense to write a copy to disk to bridge between restarts - use the native modules for requests and cryptography hashing

In my experience most of computer science is about being clever.

ubiquitous artificial general intelligence is for me the obvious contemporary north star.

the problem is that an approach to AGI has yet to be discovered so any content regarding implementing AGI from current literature should be allowed into the conscience with caution.

the solution is to recognize that however implemented the AGI mind will mirror our human mind insomuch as it will be created by other human minds

Therefor a productive question is: how does my mind work? and investigate this question sincerely using your inner voice to map the properties of a thought

then you can endeavor to confidently implement an AGI knowing that whatever content you encounter was developed from an environment similar to the one your inner voice finds itself in.

The key insight to acquire is that the quality of your system will be tied to the quality of inner voice dialogue you have with your self not the quality of books/papers published by others

Artificial General Intelligence is creating the first group of multi-trillionaires

Elon Musk & Andrej Karpathy https://www.youtube.com/watch?v=oBklltKXtDE

Elon Musk & Max Hodak https://www.youtube.com/watch?v=Wqb7P3u5ujw

Larry Page & Sergey brin

Bill Gates https://youtu.be/s7O3oCWZgjE?t=225

Mark Zuckerberg & Abhinav Gupta https://www.youtube.com/watch?v=-swj_rj3luE https://research.fb.com/publications/canonical-surface-mappi...

Jason Kelly https://www.youtube.com/watch?v=SaVsc65iu6E

John & Patrick Collison https://youtu.be/_4t12rS6_jo?t=700 ------------------------ To make a trillion

- sell ~33,333,333 cars at $30,000 - sell ~2,000,000,000 devices at $500 - sell ~10,000,000 virtual developers at $100,000 / year