Bleeding edge tech

alphacyberranger@lemmy.world to Programmer Humor@programming.dev – 589 points –
42

but the I in AI it's actually a lowrcase L, so it's short for Algorithm

It kinda annoys me that the lowercase L glyph is taller than capital A. I don't mind there being a difference, but cap-height should be taller than lowercase letters.

Illuminati

The difference being?

I'm guessing he's saying companies are still using the same human written code, but since AI is sexy right now and is being used to describe even simple programming logic, everything is "powered by AI"

That was true like 5 years ago, but now companies are just irresponsibly calling out to LLMs as a function without proper safe guards instead.

Even more likely is that AI’s that write code are trained on human created code. So they aren’t coming up with new, novel ideas to problems in most cases, they are just a far more advanced “copy and paste from StackOverflow”

“copy and paste from StackOverflow”

I feel violated

Hey just remember the classic Quora answer:

https://www.quora.com/Why-should-I-hire-a-software-engineer-if-I-can-just-copy-and-paste-code-from-Stack-Overflow

They are paying $100,000. $1 to copy and paste code from stack overflow, and $99,999 to know where and when to paste the code and how to make it work.

Domain knowledge is real, and AI might level that up, but you’ll be hard pressed to find a junior engineer armed with the same tools as a senior engineer that gets dropped into a gig and can properly utilize AI or even StackOverflow to be on the same playing field. AI can write me a function. But to figure how broken a legacy codebase is and how that function can solve an issue is why engineers are still valuable…for now

I've heard this talk where I work. Senior plebs describing things that are obviously algorithms as AI. And this of course means we had AI before it was cool.

Nothing new here. Buzzwords are the only thing senior managers can understand.

1 more...

That's exactly the point. It's just how companies market their products nowadays.

I mean, true AI isn't really a thing yet. People have been using AI wrong for a very long time now. Even ChatGPT isn't real AI.

Nobody can seem to consistently define what ai even means

Inevitable. AI is Artificial Intelligence. Nobody can define intelligence, so how can they define an artificial variety?

2 more...

It's a "gaps" problem.

Creationism has the "god of the gaps" where every new fossil forces them to set the goalposts closer together.

The people who think that human intelligence is something special have to adjust the spacing on the goalposts every time a corvid solves a new problem and every time someone figures out how to make a computer do something new.

2 more...

ChatGPT is built upon a GPT language model, which is a type of Artificial Intelligence.

(This isn't my opinion, just saying what I think they are)

They are saying it's not intelligent in any way though. It sees a bunch of words as numbers and spits out some new numbers that the prediction algorithm creates.

What you're thinking of as AI is actually a narrower version, while true intelligence is termed AGI.

Explanation:
The term 'AI' (Artificial Intelligence) refers to computer systems that can perform tasks that would typically require human intelligence, like recognizing patterns or making decisions. However, most AI systems are specialized and focused on specific tasks.

On the other hand, 'AGI' (Artificial General Intelligence) refers to a higher level of AI that possesses human-like cognitive abilities. AGI systems would be capable of understanding, learning, and applying knowledge across a wide range of tasks, much like us.

So, the distinction lies in the breadth of capabilities: AI refers to more specialized, task-focused systems, while AGI represents a more versatile and human-like intelligence.

The term ‘AI’ (Artificial Intelligence) refers to computer systems that can perform tasks that would typically require human intelligence,

That's everything computers do, though, isn't it? Pocket calculators would have fit this definition of AI in the 1970s. In the '60s, "computer" was a human job title.

Unless your pocket calculator can recognise patterns or make decisions, it doesn't fit the description.

Fair enough. What evidence have you got that it's any different than what humans do? Have you looked around? How many people can you point to that are not just regurgitating or iterating or recombining or rearranging or taking the next step?

As far as I can tell, much of what we call intelligent activity can be performed by computer software and the gaps get smaller every year.

That’s not how ChatGPT works.

GPT is an LLM that use RNN. An RNN (Recurrent neural network) is not an algorithm.

Yes, but a neural network is just a collection of ML algorithms.

Yea, but not really. The algorithms are available for free, but they don’t do anything useful by themselves. The RNN is built by training the neural net, which uses grading/classification of training data to increase or decrease millions of coefficients of a multi-layer filter. It’s the training data, the classification feedback and the processing power that actually creates the AI.

2 more...
2 more...

the computer wrote the 2nd one on accident when someone asked it to bake a cake.

3 more...

Is this part of a Hi-Lo implementation for blackjack? (Also, ewww mixed types)

Yeah, just use a char for card and test

if(card < '7') count++;
else count--;

Or something, don't mix types.

The cards should just be numbers, and an enum should be used for display names

Chars are just numbers, but yeah, an enum would work fine too, sure. The only advantage with using a char for it is that there's no conversion needed for outputting them into strings so it's a little easier. Less code, very readable, etc. Though yeah, thinking about it JQKA wouldn't be numerically in the right order which could cause issues if the program did more than just implement HiLo

I hope I'm not being stupid right now, but is that the actual algorithm for counting cards in blackjack?

Half of it. This gives you the running count. You need to also keep track of “number of decks in shoe” -“number of cards dealt since last shuffle”/52 to tell you how many decks are left in the shoe, then divide the running count by the number of decks left to give you a true count.

True count higher than 1? Start increasing your bet accordingly.