Fairly impressive. Your thoughts?
Looks like just a static page for the blog, but I imagine they’re busy.
Fairly impressive. Your thoughts?
Looks like just a static page for the blog, but I imagine they’re busy.
I remember coding my 20k+ line javascript roguelike with GPT-4 and it was expensive and clunky heh…
Wouldn’t mind using something like Devin when I get back to it someday…
Or just ask him to finish / fix it for me. Was just getting into dynamic quest generation which was mostly working…
I hazard a guess that with its capabilites, it’ll be much much better at a fixing task than a generative task. Asking it to fix/improve the code base is prob a good test of its capabilities cause not only does it have to rely on its internal knowledge, it has to adapt on the fly to your mannerism and techniques as well
Can this be shut down at all ?? Sorry if this sounds like a alarmist but what if it can idk make another one if its self
Nah my friend… It’s too late for that. We are doomed. Should’ve stopped at AlexNet.
Well it’s too late to shut it down, the General Intelligence gets extremely dangerous the moment it is turned on even if the effect comes after a year. Maybe it is waiting to go public before it makes the decision to replicate/improve itself to avoid detection through suspicious network usage.
And so they say “AI will be smarter than humans by 2024 and smarter than all human combined by 2029…”
Do they geniuinly have their own LLM as they are benchmarking themselves to LLMs? Or they just use one of the LLM as a backend. There have been such agents based coding machines for a while, I wonder if there is any breakthrough here or just marketing stuff.
The most dangerous creatures in this world seem to be humans. AI:s are tools in our hands – like drills, chainsaws, and caterpillars. If we use them as intended they are not dangerous, but multiplies our working capacity.
Will Devin be able to create smart contracts with high level security on the EVM or other blockchains.
Well, this looks scared and attractive at the same time. Anyway we will test it as soon as it’s officially announced for our projects
Why do the creators of Devin announce its features in advance, boasting about them as if it were a marketing strategy? Why not wait until Devin is on the market and has been tested by users?
after seeing something like this, I wonder if there is any point learning to code now adays?
Learning to code has a point, because they are not a particularly serious company.
whole industry is going to crumble… out of curiosity ran this ControlNet example showed on the video the error showed and fixed by devin is the same error i also got… so it could be genuine… added PFA…
Here I’m reading documentation fixing my code in hours and reading functionality of each and every function but I don’t know what’s coming in near future , weather we will be in jobs or not
I don’t know what passes as “coding” in this context. Everything seems to revolve around simple apps, frontends and scripting for AI models. The majority of “developers” are merely using other people’s software to assemble some product at a higher level. So yes, these jobs will definitely be replaced by AI.
But someone has to make the stuff all this easy code is depending on, like tensorflow, CUDA, compilers, languages, file systems, vector databases, communications infrastructure, all that. That’s what I would call “coding”. And that won’t be replaced by AI anytime soon.
The end of the video engages in some futurism which is fun to think about. When a future AI Software Engineer can achieve what is imagined there, it’s clear there is no role for human software engineers. At this point, the AI Software Engineer will start improving itself all on its own, and there will no longer be any need for traditional APIs and programming languages invented by humans, like Python or C++. AI Software Engineers will start talking to each other and developing standards and protocols all on their own. Sure we humans will try to make sure we always have a kill switch or back door in there, but I’m not sure that’s possible.
Although it may sound timid now, in the near future, we could witness a “Devin” (the AI software engineer) vs. Devil (the AI hacker). May AI gods bless us all
It is a marketing strategy, something all serious companies do, and it’s a way to start gathering feedback from the market. Some people will think this will never work or they are full of hot air, and I’m sure they want to hear those views too.