Home
Starting Up
Tech
Creators
A.I.
Lifestyle
Money
Idea Board
Vibe Coding Tools
Products
Ideas DB
Case Studies DB
Subscribe to IH+
Starting Up
Tech
A.I.
Creators
Lifestyle
Money
Join
19
Likes
0
Bookmarks
28
Comments
Report
What ChatGPT does not solve
by
Maaike
https://lucaskostka.com/posts/chatgpt
Trending on Indie Hackers
Build AI Agents & SaaS Apps Visually : Powered by Simplita ai
30 comments
The Future of Automation: Why Agents + Frontend Matter More Than Workflow Automation
24 comments
You don't need to write the same thing again
22 comments
No Install, No Cost, Just Code
20 comments
Let’s Talk: What’s Missing in Today’s App Builders?
17 comments
15 Years of Designmodo
14 comments
It feels like people focus too much on replacing existing devs, onboarding new devs, but never on empowering existing ones.
I don't say this to brag, at this point there isn't much that I can't do in web development. I can code essentially anything that I put my mind to.
But do you know what my bottleneck is? Churning out code.
Yes, I've hired people. But there is still a lot of bandwidth that comes with transfering your ideas to a manager, or writing a PRD yourself.
ChatGPT is not smarter than me, but it saves me a lot of time and so much mental energy.
Repetitive API calls, CRUD operations, boilerplate code? Not my job. Copilot solves that.
Advanced functions that I have in my head, but just don't have the energy to actually code? I describe them to ChatGPT and I get the output instantly.
Occasionally, I need to clean up the code or fix a thing here and there...
But in general, I think I'm at least 3x more productive today than before copilot & ChatGPT.
ChatGPT3 will be gone in a year. You will have GPT4, GPT5, coming up that will more than solve any of the limitations that exist today.
Regardless of this, what ChatGPT kind of AI products won't solve are basically industries that Google has not been able to replace.
Healthcare, financial advice are things that come to mind. Anything that is in the YMYL category, or needs an expert won't be replaced by AI ever.
What ChatGPT has done is taken AI viral. This is why there is a Google code red. Not because of the tech, but because of the virality. Open AI has suddenly taken over mindshare from the general population.
I tend to agree with this so much, the tech it self is basically available inside google and other companies but the fact that this is released to the public to use and the vitality it generated is actually making the buzz.
You could say that, yes. I would tend to agree with your statement.
I tend to be more optimistic about the possibility of AI 'replacing' experts/expert knowledge. I guess we are just a few iterations away from when AI-generated expert knowledge will give a big boost to the efficiency/accuracy of these services. And before we know it, the outlook will be totally different.
The thing is that AI needs to learn based on what dataset is available. This is easy for "evergreen" niches - like, cycling, or dog walking, etc. But AI cannot replace news. While it can theoretically replace doctors in the diagnosis - there is also the trust factor involved.
Would you trust a machine over a learned expert for life-altering advice? Maybe we eventually would. But I personally can't wrap my head over that for the moment.
I have had a couple of occasions when I tried doctors on a NOT so common issues.. having such experiences with doctors believe me I'm willing to take a bet on AI to figure it faster than doctors do some times (kidding 🤣 but really it will have a place I'm sure where it will help in this area massively)
Or any physical job that produces value in the world such as nurse, doctor, construction worker, electrician, plumber etc. The internet and chatbots do not make the world go around.
And we're at an important moment where businesses can get customers during the early days of new AI, it doesn't really matter better AI is coming. Those who can take advantage of the wave now will have an audience to pivot future AI products towards.
Waiting for the rest.
tl:dr
ChatGPT is a powerful tool, but has limitations in its current form
ChatGPT does not ask questions and assumes the user knows what they are asking
ChatGPT may not be helpful in identifying the root cause of a problem
ChatGPT may generate code, but users should have a basic understanding of programming to use it effectively
ChatGPT may produce suboptimal or incorrect answers if the user is unable to correctly frame the initial question
(tl:dr'd using ChatGPT)
Woaw, thank you @maaike for posting this here, I hadn't been on IH for quite some time and it is always a pleasure to come back. Insightful discussion, I really like the idea from @AustinTackaberry that it's the virality of chatGPT that is probably the most surprising.
in which countries does it work?
To extend the post's point about problem definition and answers: I've recently started writing on Medium and what I've found is that ChatGPT is pretty good at cleaning up my writing, but not that good at creating prompts and topics. Perhaps I am going about it the wrong way but at least for me the limitations seem to be not restricted to coding!
What most people overlook is that big businesses will not be built by just creating a wrapper around a 'very generic' model like ChatGPT. The products that will really make an impact will have to 'finetune' these generic models for very specific/niche use cases. That's where the money lies IMP.
It's an imperfect product like many things, but it seems unable to reproduce more than a few iterations of the same statement
I once tried asking it to write a blog by referring a link and it said it cannot browse internet. I still have to give it the whole context. But i think that too will be covered soon. But yes, you need to be smarter to get smart results.
There are Chrome extensions for ChatGPT that is able to search the web called Advanced ChatGPT. It works pretty well but does have it occasional errors.
It doesn't solve anything in the physical world.
Of course, it has limitations. We expect too much from it - it's just software at the end of the day and we are expecting human responses from it. And actually, I think we'd be more worried if it could do these things.
But tech is evolving at such a fast pace, it won't be long until AI will have the ability to do these things. I'd guess that within the next 6-12 months it will have a form of 'consciousness' that will allow it to think independently and thus come up with more intelligent answers not just based on programming/human input.
Tech has been evolving at fast pace for decades. Fast evolution is one thing but the adoption is another. In my opinion we are still at the beginning. It's hard to predict how fast the adoption of AI will be. In the end it could take longer than we think and could be used for something totally different than today.
Up
https://bapehoodie.uk/
BAPE Hoodie Shop A Bathing Ape® UK Clothing store for latest collection of Shirts, Jackets, Sweaters, BAPESTA Shoes and Bape Shorts
This comment has been voted down. Click to show.
This comment has been voted down. Click to show.