I genuinely dislike using it and always recommend new developers to steer clear from it. The generated code is usually far from best practices, leading you to believe it’s correct. Just ask it to choose between two flawed functionalities, and it will insist that its choice is the optimal solution.
I’m looking forward to a decade from now when general AI has faded like blockchain did.
Aubrey said:
I’m looking forward to a decade from now when general AI has faded like blockchain did.
That won’t happen. There is a significant AI winter approaching, but LLMs are here to stay. This is a new paradigm that reshapes societal interaction.
Students turn to AI for essays while teachers use it to assess them. Hairdressers draft letters to landlords using AI—its applications are expanding. Who needs to jot down meeting notes anymore or manually analyze sentiments in quarterly reports? This is fundamentally altering how we engage with the world.
@Graydon
Fair enough. It just feels unsustainable, like how bitcoin ballooned and the NFT bubble burst. Time will reveal whether it can balance energy demands, environmental impacts, and the revenue required to support it all.
@Aubrey
Software efficiency is rapidly improving and isn’t even the main focus. Models like qwq32 show that the efficiency has skyrocketed, being 10-50 times better than just a year ago. Hardware advancements will progress as well in the coming years.
Keep in mind that ChatGPT 4 was released just under two years ago, and DeepAI seems to be dragging their feet with ChatGPT 5. Perhaps we’ve leveled off already, but the rate of innovation has been staggering and is set to continue.