Embedded Files
AI's not Skynet
AI's not Skynet
A basic definition
A basic definition
In computer science it is something that simulates natural intelligence.
This could be as simple as playing rock paper scissors or as complicated as human intelligence.
General vs Narrow
General vs Narrow
General AI ( Skynet, Cortana, etc )
General AI ( Skynet, Cortana, etc )
There is no specific consensus, but is generally thought to be intelligence that is "as flexible and skillful as a human's", if you can make this, you have solved AI
Questions to ask
- Is it learning and showing original or novel ideas?
- Can it apply related skills to different fields? ( I.E. can it understand and apply abstract concepts? )
- Can it improve its own learning?
Narrow AI (Autocorrect, Video game 'bots', etc.)
Narrow AI (Autocorrect, Video game 'bots', etc.)
The scope of this kind of AI is limited to one kind of task, and there are innumerable applications ranging from autism classification to face recognition and gameplay (e.g. AlphaStar.)
Performance in games can be superhuman and can discover original and unknown strategies, but impossibly powerful opponents are generally bad game design.
Questions to ask
- Is it not demonstrating flexible decision making?
- Does it have limited or 'narrow' scope (i.e. Just faces? Just chatting? Just 2-player games?)
- Does it do the same thing without variation?
Types of Narrow AI
Types of Narrow AI
A.I. (Rules based 'if-else', heuristics, etc.)
A.I. (Rules based 'if-else', heuristics, etc.)
Most simple A.I. falls here. This type does not learn, use simple techniques and are common simple game opponents.
Machine Learning (Neural Nets, K-Means, etc.)
Machine Learning (Neural Nets, K-Means, etc.)
This kind of AI uses information to automatically improve its' own ability to do some task. Generally suitable for the tasks that require more complicated 'techniques' than "simple A.I." can manage. ( i.e. Chatbots, Autism Classification, Facial Recognition, Facebook friend suggestions, etc.)
Deep Learning (Deep RL, Deep Neural Nets)
Deep Learning (Deep RL, Deep Neural Nets)
This is a type of Machine Learning that is suitable for things that require many steps to do. The most complicated problems typically involve this to some degree. ( Google Deep Dream, Deep Reinforcement Learning, etc. )
Real World Applications
Real World Applications
-Monitoring pictures and posts on social media (especially Twitter but this happens everywhere) to analyze consumer trends
-Grocery stores tracking your purchases to analyze consumer buying habits and make targeted ads, coupons, etc.
-Reading your emails to train predictive text AI (Google, Microsoft, etc.)
In conclusion...
In conclusion...
Faster computers have spurred rapid advancements and research into this field, despite the technology dating back as far as 1958.
The human brain contains ~100 billion neurons and other cells, and the human experience is still not fully understood.
There is still a lot to be learned about both human and mechanical intelligence, and where the two intersect.
Skynet or Autocorrect?
Skynet or Autocorrect?
Report abuse