ChatorDie.ai reposted this
I regularly use these AIs: - GPT 4/4o - programming, learning, planning - Grok - learning, news, fun, free speech - Gemini 1.5 - working w/ huge text - Claude 3 - natural-sounding convos - Perplexity - for deep-dive research on topics Thank you to the teams that build these!
The use of these AIs by average people sucks. 1. A revolt is brewing because Facebook replaced the search bar with a Meta bar. Now no one can find anything. 2. LinkedIn's AI sucks because it automagically writes messages unsolicited so you can appear to the recipients as a MS Word mail merge snail mail spam campaign. It takes longer to erase that tripe than write a quick message. Why bother? 3. Even a simple spell checker is irritating. 4. And don't get me started on the proliferation of chat bots on websites pretending to be a real person. I think that is fraud. I don't want to proctor a Turing test just to see if I'm chatting with a real person.
Here is one for you Lex Fridman-- does it worry you that by including those regularly in your decision making process (ignoring the question of whether it diminishes your ability to perform that function yourself over time) it bends the result over time towards the average? When we work through hard problems we make thousands of micro-decisions as part of that process. By handing your decision making over to a machine-- particularly one where you don't control or understand the data model stored within it you have no idea why its giving you the information its giving you. You close off the entire branches of the solution space because they're not included in the trained data set, and these are the branches that make you being part of the process valuable... (Not anti big autocomplete, just in search of places that its the right tool for the job and am constantly seeing people volunteering away their own agency for something that looks neat) (Also using an AI for free speech... you mean relinquishing your voice to the average? Or... hearing the average? Neither of these is free speech mate!)
Does anyone think it’s worth it to pay for multiple/all models? I use GPT for personal programming and some learning, but haven’t tried anything else yet
Have you tried Pi by Inflection AI?
I'm enjoying 4o's voice functionality at the moment. I like that I can give 4o a 3-minute long speech on what I want to know about a particular subject matter and amidst my lack of brevity and filler words, it does a great job of explaining things. I haven't needed to look at the transcript provided; successfully explaining through word of mouth alone is impressive. I wonder if they'll ever add the functionality to use diagrams, charts, and sources to convey a particular point. That would be awesome for all things engineering.
Would be great to have a higher level intelligence layer to reroute prompts to the most relevant LLM to fetch answers from.
I feel like GPT 4o just gobbled up the value that you mentioned with Claude 3 and perplexity. Grok is more of a fun toy. Final death nail will be when GPT 5 or similar is released that has all of that plus the context of Gemini. Also, OpenAI now has memories, which is wonderful for working across multiple sessions. Truth be told, all of these models are exceptional. The key is exploration and experimentation. Find what works for you and your workflow.
You should try claude.ai for coding imo it’s better then gpt 4o
Founder & CEO at Digital Investment Group (DIG)
5moWe’ll look back in 10 years or so from now at these versions of AI will be like comparing a Ford T model to a Tesla