What makes something artificially intelligent?

  • Thread starter kolleamm
  • Start date
In summary, the AI in this game will be able to respond to predetermined questions, and will not be very smart.
  • #1
kolleamm
477
44
I started working on a game for vr where I want the player to be able to interact with an AI person. Problem is...what makes a character smart? It's not super hard to teach them how to reply to specific things but I'm not exactly sure how to build from there. Any ideas?
Thanks in advance
 
Physics news on Phys.org
  • #3
From what I know at the moment most (if not all) games don't really use any AI, most things plot related are hardcoded as ADG (Acyclic Directed Graphs).

The closest thing to AI that I know about are the Radiant Story and Radiant AI used in Skyrim (compare ).

I am not sure about the details, but I remember reading about some kind of AI (that would make sense in games) that was used in a software designed for training of police forces in Israel (?).

Note, that so far no AI passed the Turing test, which means interacting with them in games will be problematic.
 
  • Like
Likes collinsmark
  • #4
kolleamm said:
I started working on a game for vr where I want the player to be able to interact with an AI person. Problem is...what makes a character smart? It's not super hard to teach them how to reply to specific things but I'm not exactly sure how to build from there. Any ideas?
Thanks in advance
With "reply to specific things" I assume you don't mean arbitrary natural language (because that can be really, really, REALLY difficult :wink:), but that you mean reply to a fixed list of predetermined responses (e.g. "Hey, how are you doing?", "What's your name" etc.). In that case you can use hardcoded techniques like Borek described above.

You could also have a look at:
and
(as a sidenote, I once helped a friend with a roleplay program (fantasy) in which he wanted to generate reasonable character names, and I implemented it using Markov chains. It turned out pretty good. First I fed the program a whole bunch of real names (historical I think), while it calculated probabilities that certain syllables was followed by other syllables, and then the probabilities was stored in a big Markov chain (in this case, a matrix). Then I simply used the Markov chain to generate brand new, very often unique, names. It was fun :smile:.)
 
Last edited:
  • #5
Thank you for your help everyone. I will probably hardcode a lot of the AI since most other things require astronomical amounts of knowledge and common sense. Although I do believe that even by hardcoding the system it is still an intelligent one and will have potential for further development. I've also noticed that physical interactions with the characters such as handshakes can also be a way of demonstrating intelligence because after all we don't learn everything we do from merely talking. Either way if you do have anymore suggestions please do let me know, this is a project that will benefit everyone.
 
  • #6
Wouldn't intelligence be the ability to retain knowledge and use that knowledge in future events? If you are trying to reconstruct intelligence, writing a code that simulates "learning" from trial and error or through its environment would be the quickest of ways. Google tried doing that by allowing the bot it made to learn based off of conversations with individuals via the internet... it did not turn out very well as the bot began spewing racial slurs and vulgarity. :oldlaugh:
 
  • #7
Lillie said:
Wouldn't intelligence be the ability to retain knowledge and use that knowledge in future events?
That merely moves the question(s) to:

What is knowledge?
and what does it mean to be able to "use" knowledge?

We
know intuitively what they mean, but intuition is not a thing easily ... algorithmizable.
 
  • Like
Likes Lillie
  • #8
DaveC426913 said:
That merely moves the question(s) to:

What is knowledge?
and what does it mean to be able to "use" knowledge?

We know intuitively what they mean, but intuition is not a thing easily ... algorithmizable.
How true is this! It seems easier said than done, but I guess it is the plight of the programmer to begin to understand how to artificially create intuition. I, sadly, do not have enough knowledge in programming to give any concrete recommendations, but perhaps a quick study on how Google was able to demonstrate artificial "intuition" may help OP better figure out the concepts of AI? The brain, in my opinion, is scarily similar to a machine so perhaps we will eventually be joined by robots.
 
  • #9
Regarding AI for games, there's a saying that game AI doesn't need to be smart, it just needs to avoid doing anything very stupid. Although in many cases it does not even achieve that goal.

If you want to know a bit more about game AI, I suggest Buckland, Programming Game AI by Example.
 
  • #11
Borek is right -- while game designers use the term "AI," they generally mean a pretty small set of techniques that are not really regarded academically as AI.

The reasons are not very exciting.

Real AI requires a lot of computing horsepower, and most video games run at 60 frames per second. There just isn't enough computing power available on a typical gaming computer to run something like a deep reinforcement learning algorithm side-by-side with top-notch graphics. Neural nets and graphics also compete for the same processor, the GPU. When game designers have to choose between jaw-dropping graphics and better AI, they generally choose the graphics.

Times will change, though. We're rapidly approaching the time when a cell-phone GPU can render photorealistic scenes at 60 fps, and we'll eventually have hardware so sophisticated that full-brain emulation is possible in real time. Maybe in one decade, and certainly in two.
 
  • Like
Likes kolleamm and Greg Bernhardt
  • #12
Any software that can learn on its own is called AI. It does not need to be taught. It just needs to be give sample outputs and inputs. It figures out what to do.
 
  • #13
This might not really apply to gaming, but rather to robots. What makes a character smart and intelligent is consciousness and the ability to predict and simulate future events and ideas.
 
  • Like
Likes kolleamm
  • #14
Ivan does make a good point , that intelligence should come from being able to predict future events. In my opinion I believe intelligence also comes from being able to interpret unrestricted grammar, meaning the AI can extract the meaning of your input even if it's not in perfect form.
 
  • Like
Likes ISamson
  • #15
chroot said:
... we'll eventually have hardware so sophisticated that full-brain emulation is possible in real time. Maybe in one decade, and certainly in two.
I wouldn't be so sure.
We could well have sufficiently powerful hardware to do that,
but as yet we don't understand much about how the human brain processes signals,
or how this gives rise to self awareness and other features of intelligence.
You can't emulate something if you don't know the details of what you are aiming to emulate.
 
  • #16
rootone said:
as yet we don't understand much about how the human brain processes signals

Agreed.

You can't emulate something if you don't know the details of what you are aiming to emulate.

Agreed.

or how this gives rise to self awareness and other features of intelligence.

That's the part I disagree with. These are most likely emergent phenomena, so we don't need to put them into the simulation. To use an analogy, you don't need to put elliptical orbits into the n-body simulation of the Solar system, motion equations and initial conditions (masses, velocities) are sufficient.
 
  • #17
kolleamm said:
Ivan does make a good point , that intelligence should come from being able to predict future events.

Define "predict". Despite being a member of an intelligent species you have no idea what will be the USD/EUR exchange rate in 12 months.

Actually defining intelligence is in itself a difficult problem, so far we don't have a definition that everyone agrees on.
 
  • #18
For this thread I would define this type of intelligence of game characters as the ability to pass the turing test.
 
  • #19
kolleamm, if your goal is to make in-game characters that pass the Turing test, via written language, you're going to have a pretty hard time of it.
 
  • Like
Likes Greg Bernhardt
  • #20
Borek said:
Define "predict".

The ability to simulate future events and work towards or against the outcome.
 
  • #21
kolleamm said:
For this thread I would define this type of intelligence of game characters as the ability to pass the turing test.

Predicting future is not part of the Turing test.
 
  • #22
Borek said:
Define "predict".

Humans are intelligent beings which are able to infer on someone's actions what they are going to do next and act upon that understanding the best way to, for example, prevent that.
This is what is meant by predict.
 

1. What is the difference between artificial intelligence and human intelligence?

Artificial intelligence refers to the ability of a machine or computer program to think and learn in a similar way to humans, while human intelligence is the natural cognitive ability possessed by humans. AI is programmed by humans and relies on data and algorithms, while human intelligence is a result of complex biological processes.

2. How do you measure the intelligence of an artificial intelligence system?

There are various metrics used to measure the intelligence of AI systems, including accuracy, speed, and the ability to adapt and learn from new data. Another measure is the Turing test, where a human evaluator communicates with both a human and a machine and tries to determine which is which.

3. Can artificial intelligence become smarter than humans?

While AI systems can perform certain tasks more efficiently and accurately than humans, it is unlikely that they will become smarter than humans in the near future. AI is limited by the data and algorithms it is programmed with, and it does not possess the same emotional and creative abilities as humans.

4. What are the ethical concerns surrounding artificial intelligence?

There are several ethical concerns surrounding AI, including the potential for job displacement, biases in AI decision-making, and the potential for AI to be used for malicious purposes. There are ongoing discussions and debates about how to address these concerns and ensure responsible development and use of AI.

5. Can artificial intelligence have emotions?

Currently, AI is not capable of experiencing emotions in the same way that humans do. However, there are efforts to develop emotional AI, which can simulate emotions and respond to human emotions. This technology is still in its early stages and raises ethical questions about the relationship between humans and emotional machines.

Similar threads

  • General Discussion
Replies
10
Views
870
  • Computing and Technology
Replies
1
Views
274
Replies
9
Views
977
  • General Discussion
Replies
33
Views
2K
Replies
1
Views
872
  • Computing and Technology
3
Replies
99
Views
5K
  • General Discussion
Replies
26
Views
4K
  • General Discussion
Replies
12
Views
1K
  • Computing and Technology
Replies
34
Views
4K
  • STEM Educators and Teaching
Replies
4
Views
2K
Back
Top