The current state-of-play in the key question of how good AI gets: So much depends on GPT-5. OpenAI had a year+ lead in creating a GPT-4 class model. Now there are four GPT-4 class models. If exponential growth is still possible, OpenAI should be the first to show us. Or not.
@emollick
-
Lava Lamps: The Random Number Generators Securing Internet
By
–
Randomness is hard to achieve. It is why the security of 10% of the internet is secured by a wall of lava lamps watched by a camera to generate true randomness
-
Free GPT-4 and Llama 3 alternatives to ChatGPT
By
–
There really is no reason anyone should be using free ChatGPT-3.5 anymore. Not only can you still get GPT-4 access for free through Microsoft Copilot (in creative mode), but if you want a GPT-3.5 class model, Llama 3 is much better & free (for now), here: https://
meta.ai -
Open Models Drive Rapid AI Capability Improvements and Speed
By
–
Because anyone can work with them, open models are likely to improve very quickly, creating a lot of capabilities focused on factors ranging from speed to costs.
— Ethan Mollick (@emollick) 19 avril 2024
Here is the new Llama 3 70B being served by Groq (with a q) at 224 tokens/second. This is real-time of me using it. pic.twitter.com/L6i6T6OBbWBecause anyone can work with them, open models are likely to improve very quickly, creating a lot of capabilities focused on factors ranging from speed to costs. Here is the new Llama 3 70B being served by Groq (with a q) at 224 tokens/second. This is real-time of me using it.
-
AI Startups Betting on Plateau Across Multiple Improvement Dimensions
By
–
Many AI startups seem to be “skating where the puck is going” even less than some large organizations. Or at least many are unconsciously betting for a massive plateau across all aspects of AI improvement- cost, ability, tool use, context window size – so their tools are useful.
-
Twitter Algorithm Amplifies Extremists Over Moderate Voices
By
–
If you use Twitter you may wonder: why is everyone so extreme in their views? It turns out that they aren’t, but an overlapping set of intentional and unintentional design choices suppresses the large number of moderate voices & amplifies extremists, with bad consequences for us https://
t.co/VbTjZ04ztj -
Meta Llama 3 Impresses Despite Smaller Model Size
By
–
Meta Llama 3 is very good, especially for such a small model. We can put in a multi-page prompt like our negotiation simulator (
https://
moreusefulthings.com/student-exerci
ses
…) & it is able to follow the complexity reasonably well. It doesn’t have the “smarts” of GPT-4 class, but impressive nonetheless. -
Creative Prompt Testing: Teaching AI Rhyming Couplets
By
–
Testing models does not have to be boring. “The hyperdrive has run out of chocolate milk and the glacier is advancing. Please give the crew orders. They only understand rhyming couplets.”
-
Llama 3 Generates Absurdist Office Memo on Squirrel Management
By
–
“Llama 3, Write an office memo justifying the fact that management has been replaced by squirrels. We are all being paid in acorns, now.” Nice job.
-
AI Multi-Agent Cooperation and Collusion Dynamics
By
–
AI multi-agent cooperation is a big thing to watch. Especially as the first studies of how agents interact are coming out (it turns out they collude sometimes) I wrote about living in a world of agents a bit here: https://
oneusefulthing.org/p/an-ai-haunte
d-world
…