I just don't think for what I do that ChatGPT etc is there yet.
By making a stand and not using it won’t really help though will it, it’s happening either way, so might as well take advantage.
Its not ‘there’ for my job either (it definitely can’t write a half decent lesson plan) but it’s a brilliant tool to help get a lot of projects started.
Anything you want an answer to. I asked for the full explanation to the Israel/Arab conflict for example.
how do you know the answer was correct ?
Well yes, but how do you know encyclopedias, google or anything you read is correct?
ChatGPT literally makes shit up. For professional purposes I asked it for a list of TV competition shows highlighting excellence in retail. Some of the answers looked suspicious. i challenged it one by one, and in each case it backed down and apologised for a “mistake”. I have the transcript, it’s hilarious but too long for a post here.
Have you not noticed the caveat (warning) it now posts tot he effect that it may make mistakes (or make shit up) ?
Yes, hallucinations exist. Part of the art of what I teach others is how to construct prompts to maximise correct outcomes but also to question itself ( as you did).
As per my examples I posted earlier, none of the best use cases relate to something that can be googled and solved in 30 seconds, especially recent content. I personally don’t think this is what the tool is for. It is for complex tasks for which you have an idea, but not the skill, to execute.
Research, analysis, training seminar outlines, contents for guidance papers … and loads more
So basically the theme I am getting is people are using it to take shortcuts in their jobs.
Seems insane to me and a path to A.I replacing those jobs.
You see it as a shortcut, I see it as a facilitator.
What could take many hours of research can now be done in a fraction of the time, giving me the freedom to spend more time on the real work.
It only gives outputs to my input … if I didn’t provide the parameters, it would not be of value.
To an extent, it does feel like ‘cheating’ but it is only producing data in response to my detailed input. There can be issues with the output but that just requires care and attention.
I am now able to do my work way quicker than I could before giving me more time to develop and implement the results.
Anything you want an answer to. I asked for the full explanation to the Israel/Arab conflict for example.
how do you know the answer was correct ?
Well yes, but how do you know encyclopedias, google or anything you read is correct?
Fair point, but humans that make too much stuff up or contradict themselves would hopefully get banned from updating Wikipedia.
ChatGPT's answer above to my Ronaldo/Rooney question has Ronaldo being sent off AND scoring in the penalty shootout.
I can see uses for ChatGPT in summarizing, drafting and brainstorming.
What irks me is calling it Generative Artificial Intelligence. I think this is partly marketing, since the more specific (and accurate in my opinion) 'Generative Language Model' has nowhere near the mistery.
PWR but I suspect the Air Canada AI Chatbot issue where it basically made up its own more generous policy than the one AC had and the carrier has to honour it will put a brake on adoption of those systems for business purposes.
Much like Dalle-3, this may be useful in some circumstances, but there will be limitations. You won’t be able to edit or fix specific elements you don’t like (even re promoting won’t give you exactly what you want), you won’t be able to alter angles or colours. Anything that has words on it will be a mess.
The promo video looks good and could be useful for small clips or corporate videos for websites but SORA is not going to replace human creativity and even short films (much like Dalle-3 has not replace creative depts).
The tech will get better so who knows what will happen in 5-10 years.
One point of note on potential fake news generation (which this could be used for). OpenAI have committed to adding a watermark (C2PA) to content generated via SORA. This has already been applied to DALLE-3 images since feb. It should make it easier to identify ‘real’ images for both new outlets and individuals alike. A standard such as this needs to be adopted across the whole AI industry to be effective.
PWR but I suspect the Air Canada AI Chatbot issue where it basically made up its own more generous policy than the one AC had and the carrier has to honour it will put a brake on adoption of those systems for business purposes.
Have I just been Lol’d by an AI chatbot that’s realised the games up ?
Very interesting podcast by the Guardian called Blackbox - it has nothing to do with the Charlton recruitment process, but is about the development of AI.
Episode 1 Looks at the initial development.
Episode 2 looks at how deep fake is adversely affecting people already.
Very interesting podcast by the Guardian called Blackbox - it has nothing to do with the Charlton recruitment process, but is about the development of AI.
Episode 1 Looks at the initial development.
Episode 2 looks at how deep fake is adversely affecting people already.
Interesting first episode on the history of neural networks.
Second episode describes great detective work to track the creators of ClothOff.
But I do wish podcast transcripts were made available as a matter of course for people that haven't got time to listen to them.
I still think there's a lot of marketing in the term Artificial Intelligence. Not long ago it was all about algorithms, and just before that 'big data'.
There's so much non-intelligent technology called AI that they've had to come up with another term, Artificial General Intelligence, to excuse mistakes in regular AI.
Comments
Its not ‘there’ for my job either (it definitely can’t write a half decent lesson plan) but it’s a brilliant tool to help get a lot of projects started.
ChatGPT's answer above to my Ronaldo/Rooney question has Ronaldo being sent off AND scoring in the penalty shootout.
I can see uses for ChatGPT in summarizing, drafting and brainstorming.
What irks me is calling it Generative Artificial Intelligence. I think this is partly marketing, since the more specific (and accurate in my opinion) 'Generative Language Model' has nowhere near the mistery.
https://www.wired.com/story/google-rebrands-ai-chatbot-gemini/
But yes, I can see the potential for misuse.
Episode 1 Looks at the initial development.
Episode 2 looks at how deep fake is adversely affecting people already.
https://www.theguardian.com/technology/series/blackbox
Second episode describes great detective work to track the creators of ClothOff.
But I do wish podcast transcripts were made available as a matter of course for people that haven't got time to listen to them.
I still think there's a lot of marketing in the term Artificial Intelligence. Not long ago it was all about algorithms, and just before that 'big data'.
There's so much non-intelligent technology called AI that they've had to come up with another term, Artificial General Intelligence, to excuse mistakes in regular AI.
https://x.com/googledeepmind/status/1821562365931855970?s=46&t=ynww82GMl7VKBjthBflU0g