GPT-4O VS Claude 3.5 Sonnet VS Gemini 1.5 Pro. Which is the KING of AI?
Вставка
- Опубліковано 19 вер 2024
- Testing the three major AI models in the market on three key areas. Coding, Math, and logical reasoning.
Welcome to Everything AI where I talk about artificial intelligence. I give you all the AI news as they come and dive deeper into the research world to bring you the next big thing in AI before it happens. Consider subscribing, liking the video, and sharing your thoughts via the comments section.
🤣🤣🤣 "This is where some of the models would smoke some weed" got me 🤣🤣🤣
And it got better and better as it went on 🤣🤣🤣
As as one more person stated in the comment, you didn’t use Gemini 1.5 pro/ Gemini advanced, rather you used Gemini 1.0 pro or something that is not advanced, so make sure you use the right version of Gemini next time
I noticed this switch by Gemini about a week ago where they just decided they would only give very short answers. They used to be extremely pedantic where I'd have to tell it over and over not to
A.) Give unnecessary extra info.
B.) Stop giving me life lessons.
C.) Stop asking if me I really wanted the answer to some 'potentially sensitive history....'.
Looks like I wasn't the only one, and now they decided on some passive aggressive changes ha ha ha.
Also noticed that too.
8:40. The farmer should take the dog across the river leaving Himself behind. What reasoning is that from Gemini?
There is an evident lack of logical reasoning in AI models. No matter how good they are, they are still AI models.
where's Gemini advanced
I think the models did not consider the farmer as a lady because your prompt included himself which signifies male. But the answer of encouraging the dog to swim is funny though.
brooo
Artificial Intelligence will remain to be just a tool. They cannot get any consciousness. Logical reasoning is still a problem. Am laughing seeing the responses that gemini gave. Very Poor
I'd like to push back on that for the sake of a friendly argument with a fellow stranger on the internet. If I may?
The distinction between tool and consciousness through the metric of 'logical reasoning' won't hold against time. It is inevitable that ML software will outwit humans in logic.
We know what tool means. So let's debate consciousness. How do we decide what that is? Does it have to be self aware, feel pain, or what levels of autonomic and cognitive functions does it have to pass? So lets say logic past high school level for autonomic and cognitive function, then we're left with self aware and pain. Pain can be interpreted as physical or emotional. The methods of ML training involve punishment and rewards, if we wanted, we can also attach it to its own autonomous physical body which would than react with a sensory called pain.
Now we're left with 'self awareness'. At what level of neural network complexity do we believe the output to be intellectually autonomous in which we consider 'self awareness'? Must it be able to feel the need to proliferate? If so, how long until an 'AI' starts multiplying itself for the sake of safety, also, if we are actively proliferating it, are we not then part of its reproductive organs?
And then we have to ask, how does this all compare to humans and animals we have today? How many have the logic, autonomy, pain and cognitive functions within the metrics we created? For example, how 'conscious' is someone in a coma? Or how about a tree?
We will come to the point that these definitions will be argued by the AI itself. And then we'd be asking ourselves, ... "are we conscious?"
Are you awake my friend? Or is this all a dream?
@@littlellama8405 Great thoughts there. However, I would like to throw in some thoughts also regarding consciousness. Is consciousness biological or computational? Alan Turing argued in his paper "The chemical basis of morphogenesis" that biology is mathematical and perhaps this would link the two camps of mathematical and biological computation. If possible, then machines will be self aware and as a result be termed as conscious. My two cents.
I stand on the side of biological consciousness rather than the computational consciousness. AI will remain to be geometric transformations of of encoded data. Simple. Nothing more. They may show signs of intelligence but Not truly intelligent.
@@BeautifulMotherNature I agree. As expanded by Y.N Harari on algorithms as part of our consciousness.
About 10 years ago I was at a meeting on Autonomous Machine Learning (AI) 'good or bad?'
oddly enough, almost every one voted bad. This was a meeting of about 200 people with the UN and some deep pockets included. I argued an agnostic approach, much like having a child, the best we can hope is that we do a good job in a good environment. Everyone has the capacity to do great evil, yet the vast majority of humanity is altruistic in a holistic sense. I do understand that this can be argued as 'fear mechanisms of society' that create a 'stable society'. (Much like how the congregation walks over the homeless on their way to a grand church to put cash into a golden box). But we seem to be doing Ok at the very least, with 100% of our major wars a lack of communication which brings a lack of empathy.
AGI may not have that lack of communication. And by that sense a simple algorithm provides it with empathy. And if that isn't consciousness then what is? :)