Last I heard Google hadn't gotten this to a stage where they were willing to do a live demo, so I'd be unsure how far along it actually is.Some of the enhancements are so impressive. Little things like inflection and utterances like 'er' make it sound so much more natural, and some of the bigger things like negotiating times and dealing with idiots who think 7pm means 7 people is really impressive. Apparently if the AI can't come to a definitive result it will pass it on to a human at Google for them to complete so that they can then use their own responses to improve the AI further.
It's apparantly coming very soon. They have thousands of simulated conversations it has run.Last I heard Google hadn't gotten this to a stage where they were willing to do a live demo, so I'd be unsure how far along it actually is.
It's still AI if a neural net is making decisions around appropriate responses. Apparently it even takes the tone of the replies into account.I'd also consider it a UI and not an AI as it's working from preprogrammed responses.
It's only an AI if it's self aware. A UI is perfectly capable of selecting an appropriate option from a list of responses using preset parameters, what it's not capable of is creating it's own responses when faced with a situation it has no experience of, hence it defaulting back to a human when necessary.It's apparantly coming very soon. They have thousands of simulated conversations it has run.
It's still AI if a neural net is making decisions around appropriate responses. Apparently it even takes the tone of the replies into account.
Your definition of AI is clearly different to the rest of the industry then. You are one of the few who thinks AI requires self awareness. AI as generally referred to is applied machine learning and little more.It's only an AI if it's self aware. A UI is perfectly capable of selecting an appropriate option from a list of responses using preset parameters, what it's not capable of is creating it's own responses when faced with a situation it has no experience of, hence it defaulting back to a human when necessary.
Does that make it any less impressive? Their AI held a discussion with someone and it managed to understand it didn't need to make a table reservation despite the fact they misunderstood what it was saying a couple of times and had a confusing accent.It's coming very soon to a small group of beta testers and there's a huge difference between simulated conversations and a live demo in uncontrolled conditions.
Machine Learning is one small subset of AI research. Speech recognition is far from the be all and end all of human like behaviour and intelligence.Your definition of AI is clearly different to the rest of the industry then. You are one of the few who thinks AI requires self awareness. AI as generally referred to is applied machine learning and little more.
It's a slight improvement on the likes of Siri, Alexis and Assistant 1 then.Does that make it any less impressive? Their AI held a discussion with someone and it managed to understand it didn't need to make a table reservation despite the fact they misunderstood what it was saying a couple of times and had a confusing accent.
Almost all of which are varying levels of applied machine learning. You accept that the industry considers these specific applications to be AI then? I mean Google even refer to it as AI. My own work includes work with shallow neural nets and decision trees amongst other algorithms, and I would refer to those elements of what I work on as AI.Machine Learning is one small subset of AI research. Speech recognition is far from the be all and end all of human like behaviour and intelligence.
I would say more than 'slight'. It actually holds a conversation with people and understands the context of their comments even when those comments are unclear. Compared to current tech, and given the steps needed to move it on to the next level, I find it deeply impressive.It's a slight improvement on the likes of Siri, Alexis and Assistant 1 then.
As I said, why haven't they been confident enough to do a live demo?
It isn't restructuring it's own thinking though, is it? It's defaulting back to a human who is using the experience to add further programming. It's a UI.Scientists have achieved the following:
I don't think it takes much of an extension beyond that where an AI could experiment with input parameters to produce unconventional and unique AI. What does 'waking up' mean? An AI that can restructure its own thinking based on experience? Is that not what such an AI would have achieved?
- Produced AI that can experiment with scenarios to produce unconventional and unique solutions to specific problems
- Produced AI that can structure and execute other simple AI to interpret specific input parameters
In your opinion. The industry calls it AI and arguing about it is nothing more than an argument about pointless semantics. Call it what you like, you now know what the thread title refers to.It isn't restructuring it's own thinking though, is it? It's defaulting back to a human who is using the experience to add further programming. It's a UI.
I didn't say it is, I said the self-reprogrammability is what I think is the main missing link to 'self-awareness'. The building blocks are mostly there, but the links between are missing.When it can patch itself come back.
That's precisely what I mean. If you get a neural network that can create another neural network (done already) and a neural network that is complex enough to make novel and unpredictable decisions based on predictable inputs by experimenting on solutions and testing the answers (done already) then how much is left to do before those neural networks can start to experiment with neural networks to achieve the intended outcome? If it knows the required result, produces a neural network that achieves it and replaces its own logic then would it not also most likely ultimately become self-aware?'Waking up' I've took to mean, is a machine acquiring consciousness. A subject of experience who 'chooses' what to do. Everything created so far has been programmed, the computer doesn't actually 'think'
This could change in the future, as we're just biological robots ourselves.. but our brains and perception etc is something a computer or machine is waaaaay off understanding/experiencing.
Get a certain amount of neurons firing in such a way and consciousness emerges.. that's my belief, and I think we're miles away from doing that mechanically.