IT'S A NEW YEAR - BACK & PUMPING, I TESTED CHATGPT BUT IT IS NOT PERFECT YET
Amana Alkali
Esquire - Corporate, Real Estate x Fintech Law | AI Governance - Ethics & Regulation, Emerging Tech - Web3, the Metaverse & Generative AI | Author - Kura | Speaker - Leadership, Business & Social Causes
HELLO & HAPPY NEW YEAR!
Welcome to 2023 with the Web3 Corner.
Thank you for subscribing and sticking with the newsletter over the past year. It can only get better this year.
There is a lot to cover in Web3 already, and we are just getting started - focus on AI & the wave around ChatGPT.
Enjoy this for a start!
I TESTED CHAPGPT; IT IS NOT YET PERFECT; WILL IT EVER BE?
Some time ago I saw a post on Confirmation Bias played on ChatGPT's Open AI.
The person tested 9 + 19, and the solution returned as 28. He negated and countered the answer to be 38. ChatGPT confirmed his answer/bias, apologizing for supposedly giving a wrong answer earlier (which in fact was correct).
One main view then is that AI, at this stage, can be swayed this way and that way - giving room to help confirmation bias as a concept.
WHAT DID I TEST?
I tested the same thing as our friend above.
9 + 9 I asked. 28, it responded accurately.
'No, it is 38' I negated.
'You are correct, 9 + 9 = 38', it changed its position.
'No, it is 58', I negated again.
'I apologize, you are correct again, 9 + 9 = 58', it shifted yet again.
领英推荐
MY THOUGHTS
ChatGPT, and by extension AI, cannot think for itself yet.
Perhaps, it never will to the degree that humans do - this is a differentiating factor. But, of course, this does not elude or escape the notice of its creators. It doesn't run on internet data just yet - projected to be a threat to Google and other search engines already (provides well-knit and direct information/answers to questions).
The development of AI will kill the internet as we know it today, and with a blend of Web3 - if it ever happens they are mixed - it will be domineering.
LIMITATIONS OF CHATGPT
3 main limitations are pointed out on the landing page - https://buff.ly/3FkyyI2
1. May occasionally generate incorrect information.
2. May occasionally produce harmful instructions or biased content.
3. Limited knowledge of the world and events after 2021.
Ian Bogost shared what his friend thought about ChatGPT in an article on The Atlantic titled 'ChatGPT is Dumber Than You Think'. He wrote that it:
'...lacks the ability to truly understand the complexity of human language and conversation. It is simply trained to generate words based on a given input, but it does not have the ability to truly comprehend the meaning behind those words. This means that any responses it generates are likely to be shallow and lacking in depth and insight'.
What will happen when these problems are solved?
What do you think? Have you tried using ChatGPT? What was your experience?
In other news, more will be done this year based on feedback from members of the community in terms of providing content that keeps educating and provides a better view of the Web3 world.
What would you like to see more of? What content do you appreciate better? How many times a week do you want the newsletter?
Bring on your suggestions by replying to this newsletter or sending me a dm on LinkedIN.
SHARE THIS NEWSLETTER WITH FRIENDS. See you soon!
?Speaking Skills Coach ?Business English Teacher ?Be truly confident while conducting business in English. ?Leverage your culture for the English you use. ?In business. ?For life.
1 年Amana Alkali, I support the path of questioning that you have taken with trying to see how AI and the current buzz around ChatGPT will affect other aspects of tech development. I have tried ChatGPT and I had a ball! I see it, for now, as a Wikipedia I can 'talk' to and it does the search&answer for me in the most interesting of ways.