ChatGPT is like a very stupid and untrustworthy intern - it is not taking over the world!

in #chatgpt10 months ago

So I used ChatGPT for the first time last night.

I was very underwhelmed.

I asked it to do what I thought was a very simple task (suited to a computer), preparing a vegetarian / vegan meal plan and shopping list for 12 people for 12 days on a boat at sea without an oven.

This is something fairly easy to do by searching the internet for recipes and putting the ingredient lists into a spreadsheet for aggregation.

It did OK on the actual recipes but was really terrible on the maths of ensuring recipes were for 12 people and on aggregating all the ingredients into a shopping list.
It also failed to ensure that there were vegan and non-vegan options for each meal as requested.

Despite numerous efforts to get it to correct these errors I could not get it to produce a useful result that I could trust.

Every iteration would result in some things fixed but new errors introduced.
I could never be confident in the results. It is no good if you find out 100 miles from land that it forgot or under-provisioned key ingredients!

The experience was like dealing with a really stupid intern that was also untrustworthy and you felt was lying to you or just saying what it thought you wanted to hear.

It was not worth using it for the task and if you wanted a reliable result you needed to do it yourself.

I was very surprised that a computer would be so bad a maths.

So my conclusion is that AI is a long way from taking over the world and while some people will lose their jobs to AI, they are people doing stupid, boring jobs that are unworthy of human intelligence.

These people will be better off and happier if they are forced to use the truely unique qualities of human beings, the ability to innovate, think outside the box and empathise.

Screen Shot 2023-04-14 at 11.55.25.png

Please vote for my Hive witness. (KeyChain or HiveSigner)

Witness Vote using direct Hivesigner


just saying what it thought you wanted to hear

That's the whole point of AIs, the results they provide are statistical likelihoods, not actual reasoning, let alone knowledge or wisdom gained through experience.

We are being sold an empty lie, with the sole intent I believe, to have us willing submit our best ideas to the AI platform, so that those who control it can profit from it.

These AI platforms have already been scaled back for public use because they were deemed "too powerful" because people were using them in "unexpected" ways, meaning, ways that would subvert the world control and power paradigm.

So true. That's why I only asked it to do something I'd ask an intern to do.

I'd never involve it in something important.

Most people are being wowed by the toy-like circus tricks it can do.

Programmers are going to be out of a job? Not likely, as your task demonstrates, you still have to think like a programmer to have it perform your desired task as you wish, otherwise potential disaster ensues.

Even AGI is not going to be a magic genie, simply because you're going to have to think through very carefully what the rules and constraints are.

If we get to the point where AGI can anticipate or "mind read", then we are in trouble, because of the unintended consequences of non-specified, pre-empted actions AGI would take when it determines our "subconscious" desires.

Such activity will be enabled and driven by those who own and control the AGI, wanting to take advantage of and profit from the end user, in exactly the same manner the big social media platforms currently do. I'd go far as to say they will be the first to do it.

Did you really think they would release it to the public, if it would endanger the current order? Not released to the public, they for sure have several generations more advanced versions.
What is out there is just a toy for the masses to get amused and distracted.

Agreed, but I would take it a step further; to harvest the best and unique ideas, so again enabling the greatest wealth transfer in history, but one that includes the wealth of knowledge, because knowledge is power.

What wealth transfer? The true knowledge is not what ChatGPT spits out, just a c&p from the internet.

No, each time use ChatGPT, you are training it. It is called reinforced learning. It's similar to what you do when use a CAPTCHA (AI training) to access a website.

People are also uploading or pointing ChatGPT to new sources of information that it previously did not have access to.

So my point stands, that it is a (knowledge) wealth transfer.

Then there's the societal impact, with an estimated 70% of people being made unemployed and reliant upon Universal Basic Income. UBI certainly will not make you rich, but it will for those who control AI. Control is the theme, because if you are dependent upon UBI to live, then those who control the UBI control you.

True. Besides it is humiliating to depend entirely on others.

That's the point of it, to disempower people and turn them into willing slaves.

It is interesting, how pretty simple algorithms are called AI nowadays. Intelligence is way more than just counting and pulling information from memory, but perhaps the meaning of the word has changed.

The algorithm is as good as people who coded it and since they are deep in the woke culture, it is no surprise, if 2+2=5 in the secret codebase.

Of course you're right, AI is far from taking over the world. That was probably one of the simplest of tasks and imagining how it failed terribly at it is a proof that it's far from perfect. Those who will keep their job will keep it.

That is so weird, you would think that calculation would be easy for it.

Chatgpt is lacking when it comes to calculations. Have you tried GPT4 ?

I'm pretty sure going forward, it will surely improved in certain things,

Any public AI will be limited and throttled, because those in control do not want the current order of power subverted or overthrown. This is why in a very short amount of time ChatGPT was hobbled, because people were already asking it to "hack banks" and look for and exploit "security vulnerabilities". Do you honestly think they will let Joe Public drive the Ferarri? Nope, it will be the "city bus" version of AI for the public.

Great example when ChatGPT is failing as up until now everybody was praising it. I think that it might be useful for more technical tasks like programming as I used it to make some functions that I needed at work. Thus, I think the human is still needed to put everything together and make the logic behind things.

I agree. Look at the incentives: anyone in the media or on social media, whether they support AI or are warning against it, will tend to overestimate its capabilities in order to attract attention.

I've been enjoying Bing search chatbot. Comes standard in Edge browser. Replaced google for the most part when I need a quick answer. Still provides links to sites and it stays on topic so I can be vague as I progress. Tells terrible jokes though...

It is still more a program/algorithm than an AI, isn't it?
Maybe it's advertised as AI so people trust it (vs. it being the code & opinion of just a group of people)?

It is so sad that some people are even depending on it now
The shit is crazy

Good test! 😄