This is a great read! Thank you for the practical advice.
But to be clear, LLMs as currently constituted will simply never completely overcome the hallucination problem. They lack a coherent internal model of the world, don’t have any intuitive understanding of logic, and frequently make the wrong conclusions even when presented with factual information.
It’s an inherent limitation to even the most advanced LLMs and there is no obvious solution (other than entirely new architectures such as world models).
I recommend checking out Gary Marcus for more on this.
I came across your posts through Substack's algorithms — love this one in particular since it has a less 'intense' approach to how we should use AI. Not denying it entirely nor exploiting it, but trying to be more aware, and staying 'humane'. The last principle hit me hard as I tend to use AI to proofread or make my personal writing/message sound smoother and more articulate (perhaps since I'm not a native user of English). Perhaps from now on I'll unlearn this and try to embrace those imperfect yet sincere pieces :)
Thanks for reading, Gia! Yes, I think it boils down to intentional use at the end of the day. And the ‘right’ approach is not one-size-fits all! Personally, I think imperfections add a touch of human-ness in today’s AI world :)
This is a great read! Thank you for the practical advice.
But to be clear, LLMs as currently constituted will simply never completely overcome the hallucination problem. They lack a coherent internal model of the world, don’t have any intuitive understanding of logic, and frequently make the wrong conclusions even when presented with factual information.
It’s an inherent limitation to even the most advanced LLMs and there is no obvious solution (other than entirely new architectures such as world models).
I recommend checking out Gary Marcus for more on this.
Thanks for sharing your perspective!!
What a refreshing read. Thank you!
Thanks for reading, Ciara!
This is such great advice!!
I’ve always considered chatgpt a great tool to brainstorm with, but now am realising I am depriving myself of what makes my work and thinking unique.
Thank you for providing your framework ! I’ll be integrating this into my workflow :)
Also agree that it’s a great tool to brainstorm as long as we use it to enhance rather than replace. Thanks for reading, Angelique!
I think I will have come back to this many times in the future
Thanks for reading, Lenina. Great to have you here!
I came across your posts through Substack's algorithms — love this one in particular since it has a less 'intense' approach to how we should use AI. Not denying it entirely nor exploiting it, but trying to be more aware, and staying 'humane'. The last principle hit me hard as I tend to use AI to proofread or make my personal writing/message sound smoother and more articulate (perhaps since I'm not a native user of English). Perhaps from now on I'll unlearn this and try to embrace those imperfect yet sincere pieces :)
Again, thank you for the wonderful post!
Thanks for reading, Gia! Yes, I think it boils down to intentional use at the end of the day. And the ‘right’ approach is not one-size-fits all! Personally, I think imperfections add a touch of human-ness in today’s AI world :)
Great way to contemplate our relationship with AI as we continue to inject more and more of it into our lives 💙 love your honesty, Ines!
Thank you, Becky!