So I tried Asking DTF related question to AI

insurgentx said:

How about if you see it as a tool? not Artificial "Intelligence", just a tool. I personally think it's a slightly better google, like you can type a very specific question and the answer would appear instantly instead of having to comb through multiple search results which some of them can be irrelevant (such as promoted content)?

Click to expand...

It's a tool, yes, but you have to constantly check that the answers it provides are correct. It's absolutely hopeless at maths. It will outline a method for finding the result of a mathematical procedure alright, but then gets the wrong answers. I've often asked it to correct itself 3 or 4 times and it still comes up with the wrong answer.

They say AI is great at writing code. It does seem to do a decent job, at first glance, but fails to understand things like storing API keys server-side to avoid revealing them to anyone who simply right-clicks a page to view the source code.

At the end of the day, ChatGPT4 is just a glorified search engine ,like Google, Bing, etc that relies on data fed to it by the OpenAI training program., and is in no way 'intelligent.'

I wrote a Connect 4 game years ago for my computer (an Enterprise 128 believe it or not! Anyone remember them?) in Z80 machine code using the MiniMax Algorithm and it performed so well I had great difficulty beating it myself! However, I would never class such a program as intelligent.

All of that aside, for those who believe AI can be creative, consider the Go competition between Korean Lee Sedol (World Champion, 9th Dan) and Deep Mind's AlphaGo AI system.

AlphaGo beat Lee 4:1 out of 5 games. There was one particular move made by AlphaGo in the second match (move number 37) that completely blew the minds of all the experts watching as well as Lee himself and even the AlphaGo team ...


Amazing!

insurgentx said:

How about if you see it as a tool?

Click to expand...

A language tool that is.
It can generate well written text, but it cannot give you the solution to a challenging technical issue.
It can also generate generic responses, which are perfect for writing useless articles.
1. Adjust blah-blah-blah ,
2. Optimize blah-blah-blah ,
3. Use high quality...blah-blah-blah,
etc...

webtrekker said:

They say AI is great at writing code. It does seem to do a decent job, at first glance, but fails to understand things like storing API keys server-side to avoid revealing them to anyone who simply right-clicks a page to view the source code.

Click to expand...

People hyping things up for their YouTube videos? You don't say!... 🤣
I agree of course, and ChatGPT is really good for increasing productivity, in some cases.
It can do basic things in popular programming languages and frameworks.
When asked to make a technical suggestion however, the response will often be either sub-optimal or totally unusable.
Not surprising of course... It is NOT cognitive intelligence.

Here is an interesting video about AI.

The relevant points form the video:
1. AI chat bots are not cognitive intelligence, and the way they work is surprisingly basic 5:32.
2. All AI chat bots have the "hallucination" flaw 9:34.
This is not real hallucination of course, but a side-effect of the previous point (1).
3. The "black box" 11:59.
Calling it a "black box" gives it a paranormal vibe, similar to "dark mater" and "dark energy" in the universe.
There is nothing paranormal in it, and what happened in the example given does actually make sense.
As mentioned in point 1, the AI is predicting the most probable next word.
If the previous words are Bengali, what language do you think the next word is likely be?

AI needs to learn before producing content. if there is some special industries which has less content contributed online (for example, advanced knowledge only known by veteran, transfers mouth to mouth or any other way than publish enough online), AI (or any kind of bots) has no resource to learn.
But when there is enough content to learn, AI could learn output valuable content.
DTF, DTG, industrial level printings are something like this.

On the other hand, the biggest difference between ChatGPT and common search engines is: search engine find valuable content for people and leading them to the source contributor which bring values (reputation, respect, money, etc.) to them, but AI learns and produces and earns for itself, the original contributor may have spent days for researching and finding and writing (or write a tip according to years of experience), but the outcome by AI brings nearly nothing to himself/herself under the process of AI bots like ChatGPT.
As a result, as far as I've heard, in the industry (AI, Internet), reputable contributors and search engine programmers have started some talks about if they still wanted to contribute content to AI grabbing, there could be technical way to avoid AI "crawl, grab" content.

Jane_D said:

AI needs to learn before producing content. if there is some special industries which has less content contributed online (for example, advanced knowledge only known by veteran, transfers mouth to mouth or any other way than publish enough online), AI (or any kind of bots) has no resource to learn.

Click to expand...

AI can be trained on private data (self driving cars for example).
It's also fairly easy to add private data into existing AI models and keep the combination private by not releasing it.

Jane_D said:

As a result, as far as I've heard, in the industry (AI, Internet), reputable contributors and search engine programmers have started some talks about if they still wanted to contribute content to AI grabbing, there could be technical way to avoid AI "crawl, grab" content.

Click to expand...

Unenforceable in most cases, because published content is published content.

webtrekker said:

Note that this isn't the same as Photoshop's 'Puppet Warp' feature. This is waaaay more advanced than that. You tell the AI where you want to move parts of your image and the AI completely redraws the image using its 'understanding' of how the subject areas should behave.

Click to expand...

Most of these are actually very similar to Photoshop's (or other image editor) 'displacement/bump map'.
Have a look at this example of 3D animating a 2D photo. No AI is involved.
It is done using the Pixi.js rendering library and a displacement map of the photo.

Existing AI technology (in-painting to be exact) takes it one step further, making it possible to generate videos by creating additional frames (in-painting the previous frame).
Image generation was the big step. The examples in the video are just implementations of it.
All very useful in the hands of the right person, but still NOT cognitive intelligence.

splathead said:

You read it here first. Top professions that will become extinct because of AI

Graphic Designer
Photographer
Architect
Interior Designer
Video Editor

Click to expand...

I would add many more professions to the list, BUT I would not use the word "extinct".

The industrial revolution had a huge impact on the jobs people were doing prior to it.
Agriculture used to provide 90% of employment. and it's now less than 2% .
People have adapted and are now doing other jobs.

The introduction of digital cameras is probably a more relevant example.
It made film drop-off shops obsolete in the early 2000s, BUT these shops are are now DTG shops or whatever.
In reality, digital cameras have made photography more affordable, and have created income opportunities for more people.

into the T said:

at 10:33 in your vid you get to the true point of ai, another thought control mechanism for the masses

Click to expand...

Not much different to the sponsored, biased, and often deceitful MSM we have now.
AI tools will just make it easier for smaller players to compete in the game.
People fear the unknown.... but adapt eventually.

TABOB said:

AI can be trained on private data (self driving cars for example).
It's also fairly easy to add private data into existing AI models and keep the combination private by not releasing it.

Click to expand...

Correct, but unless enough private datasets consecutively "imported" and always better algorithm, otherwise, AI won't be intelligent enough and less improvements can suit the always-changing-real world (e.g. until now, the prediction on traffic congestion cannot 100% make true). That's why Google is updating the algorithm to identify Ai generated content and giving lower score.

TABOB said:

Unenforceable in most cases, because published content is published content.

Click to expand...

just like Brainly is doing, it is restricting bot crawling by using cloudfare and other ways, such as limit the "agent" or only open specific server, ip, etc. to crawl the website.
the battle of attack and defence is on the way - some Ai companies or crawling software can "simulate" crawler agency and type, but the defence side is also improving to identify the simulated one or real one.

Bro, or Bob, I don't meant to bargain, just wanted to have little more discussions.