Comments by "HighTechGeek" (@High-Tech-Geek) on "AI Search" channel.

  1. The melodies in both these are pretty sloppy and all over the place. The vocal timings are often off as well as bad lyrics (too many or too few syllables) are forced into the allotted beat structure. I was hoping they would have brass, as that seems harder to get right. It's impressive and all moving forward, but we need more control over the project after it's generated so we can regenerate parts, control the melodies, harmonies, etc.
    1
  2. I've been using Pi for months. I can talk to it and it can respond in almost real time. This is an incremental step forward, reducing latency. UPDATE: Actually, they are "buying time" by having the AI sort of repeat the question back and make generic statements before actually addressing the question. The delayed answer seems more "real time".
    1
  3. My brother and I would get his Amazon Alexa and Google Home to talk to each other years ago. It was quite hilarious.
    1
  4. Elon Musk shutup because OpenAI released emails from Musk showing that he was lying.
    1
  5. I look forward to conversational AI merging with these models so we can just chat back and forth with the generators and tweak our images together into a final image. I'm sad the conversational Pi chatbot was abandoned. Hopefully OpenAI and Apple release their models to the masses.
    1
  6. Whenever I generate on Gwen's website, my GPU usage maintains 90-100% while rendering for 15-20 minutes. Are they doing the video processing locally??
    1
  7. Can you do a video explaining checkpoints, tuning, weights and models? I don't understand how these things work or are created or come together on sites like CivitAI. Thanks
    1
  8.  @theAIsearch  Awesome, I'll check it out, thanks! Good luck with the new GPU. Did you find a 4090 for $1600?
    1
  9. ​ @theAIsearch Wow! So jealous! Congrats!
    1
  10. I got Copilot to sing. It was actually very good. But after talking for a few minutes, it would simply stop responding and I'd have to close it and restart.
    1
  11.  @choppergirlfpv  How did you get the talking vs singing and the applause?
    1
  12. Apparently Copilot also has time limits. After about 30 minutes of chatting, it popped up and said "You have 3 minutes left today". Gah!
    1
  13. Businesses will pay to have the OpenAI Operators surface their products and promote them as the best deals. This is where I see this going.
    1
  14. Adobe was cloning Obamas voice with VoCo and Jordan Peele on stage over 7 years ago. Microsoft was cloning voices with only 2 or 3 seconds of audio like 2 or 3 years ago. Invideo can clone your voice today from a 30s clip. Heygen can change your voice into a different language using your voice. Not sure why OpenAI is holding back on this when it's already been around forever...
    1
  15. 0:50 6:30 These are not AI generated videos. The first on is taking existing video shot from multiple angles and putting it into a 3D environment like Blender and rendering a 3D world. The Genesis one is also just a 3D modelling program generating CG. I'm sure it's better, but they've been doing this since Jurassic Park in the 90s. They've also been doing this with corn for a long time. The 3D corn you moved around on screen is the perfect example. It's just a 3D rendered world.
    1
  16. HotShot: "Current wait times are 3 hours and 26 minutes... click here to upgrade and avoid the wait."
    1
  17. The lip syncing is pretty bad here. YouTuber Charlie Hopkinson has been doing face replacement with perfect lip sync for years and I cannot for the life of me figure out how he does it. I'm sure it's not live though and if his hands interact with his face, the hands mostly blend into the background. But I prefer what he's been doing to what's presented here (except the fact this is done live - incredible - but I don't think this would be very convincing in the real world). Can you investigate what he might be doing?
    1
  18. Any medical breakthroughs discovered through AI should be free for anyone to use. Just like we can't copyright AI generated images, AI generated medical solutions should not be allowed patents.
    1
  19. 33:02 why are there 3 different coffee shops all right next to each other?? Luckin Coffee Manner Coffee Starbucks Coffee
    1
  20. Coffee makes my mental alertness go from 40 to 80. Who knew? /s Why are you just accepting the Tesla analysis numbers as true? When I actually do the work to check AI answers like that, I find they are mostly wrong. We are doomed if we let AI teach our children.
    1
  21. ​ @samuelbucher5189 you could say similar things about ai. Before they are fed any data, they are already capable of parsing data, calculating, receiving input, generating output and all their other base functions.
    1
  22.  @inviktus1983  You missed my point entirely. Never did I say "why bother?". But our current track has 0 sentience and people think if we just put in more data and faster processing, that sentience will emerge. Sorry pal, but that ain't the way.
    1
  23.  @inviktus1983  didn't say you said it. I said it. I was clarifying my point that i made before you even chimed in. Real talk: i agree with your assessment to some degree. While the current path we are taking with AI may help us towards our journey to create sentience, it's more akin to building a hubcap on your way to creating a car.
    1
  24. Regarding "RUNNING AWAY"... I have the same issue in MidJourney just generating images of a person running away from someone or something. 99% of the time, the image renders them running TOWARDS the group or thing. Anyone have any tips for a better prompt?
    1
  25. 30:51 The Disney droids are controlled by "animators" which you can see walking behind them in the Disney clip with controllers.
    1
  26. I found a demo space that worked but I couldn't see how to prompt for different emotions... ah... Multi Style! The demo space I found doesn't offer that 😞
    1
  27. Can't use your daily free credits with Kling 1.5. Also it costs 3.5x more credits than Kling 1.0. The greed has begun to set it. For $9 a month, you can generate about 20 5-second shots. Not great.
    1
  28. ​ @risphereeditor Where can I find samples of flux styles? All I see is a few generic ones like anime or Pixar crap. MJ has 10s of thousands of diverse artistic styles, supporting centuries of art history. I haven't seen that with flux. I don't think flux is trained on that.
    1
  29. ​ @NewJeanius  Not me. But now I can. Just like a photographer using a camera to make art. Love it!
    1
  30.  @randymarsh6363  I'm as much an artist as any photographer. And you don't know my background. I'm a graphic designer and 3D artist. I just could never draw blobby natural things like realistic people or animals. Now I can, even though it takes a lot of work to get AI to do what one wants, similar to getting a good photo. There's a lot of crap to wade through, but the tools get better every day.
    1
  31. The only things Apple verified as coming to Siri this year are the glowing edge, the ability to type, the history (so Siri can reference your last few inquiries) and it knows about the "send later" feature in Messages. That's it. Everything else discussed was AFTER the statement: "Over the course of the next year, we'll be rolling out more features." She then lays out some features as possible features where Siri will be able to do x, y or z. No guarantees. Just endless possibilities. lol. There's going to be a MASSIVE disappointment when iPhone users find out that Siri can't do 90% of what was discussed in this WWDC event. It was simply a dog and pony show to help stave off Apple's further falling behind the other AI players (Microsoft, Google, Amazon, Meta, X, etc). It worked. Apple's stock rose 7%. sad.
    1
  32. Why all these cameras? I don't get the trend. Zoom and wide angle. Done. I'd rather have a high def camera on the front so I can see my selfies as I take them. That's why everyone takes selfies in the bathroom so they can see themselves in the mirror since the screen is facing away from them.
    1
  33. OMG, these things aren't sentient! They don't have feelings ffs. It was told to do a task. It evaluated that if it did certain things, it would get shut down and not be able to complete it's task. So it used ways that are part of it's dataset (i.e. lie, deceive, manipulate, sandbag, etc.) to accomplish its given goals. Do you feel sad for killing the NPCs in a video game that are following their instructions to kill you? Are the enemies in DOOM sentient?? When you use words like "it makes me sad" to analyze this scenario, it actually makes me sad that anyone would waste any sympathy on a circuit board. Save your sympathy and empathy for real people my dude.
    1
  34. In the video games, the AI can do something that always gives an expected result (with maybe a few alternative results). It can play a million years worth of this trial and error in a few days. In the real world, there are infinite results with every. single. move. And it has to play in real-time. I don't see this translating very well.
    1
  35. These sims account for ~100 elements in a world that has an infinite number of elements. Expect lots of death and failure.
    1
  36. That guy is barely touching that Unitree robot, lol. That thing reminds me of the little plastic Sony Aibo dog robots from over 25 years ago.
    1
  37. UBI? Who needs money if robots take care of our every need? Just put our brains in a jar with a continuous dopamine drip... for eternity. Bye
    1
  38. "Hey Scarlett" - bahahahahahhahahahahahahahaha!
    1
  39. Can you run tests to see how o1 handles homophones? I have yet to see a model understand them correctly. They cannot construct poems or stories or lyrics using homophones.
    1
  40. Can you queue up a whole bunch of different generations in ComfyUI so you can just let it run overnight and find the 50 or so various generations the next day?
    1
  41. 7:09 You can see the 3D effect for yourself without glasses or aids if you converge the 1st and 2nd images (ignore the 3rd image on the right) using either cross-eyed or look through techniques similar to viewing Magic Eye puzzles. Doing this, you can watch all the videos in 3D. I found the resulting depth to be very minimal.
    1
  42. 3:11 the eyes and lips don't match at all in "Ours" 13:40 the eyes don't follow the Ground Truth at all in "Ours"
    1
  43. I just opened ChatGPT to use the Advanced Voice as a free user and after about 10 minutes of chatting it stopped and said I would have to wait until NEXT MONTH to get more time. 10 minutes a month is nothing. I was mostly just trying to get the bot to do one thing and it kept reverting. Google's Gemini chat bot is a joke and can't do anything more than just basically reading text to speech. I haven't tried Copilot yet, but don't have high hopes. Maybe they DO give us access to OpenAI's bot? That would be nice.
    1
  44. What you call Monte Carlo tree search is just a brute force methodology to find answers. There's no intelligence. It simply brute forces through iterations to find successful paths. Then it presents the shortest path. Humans have some weird desire to attribute super intelligence or sentience to these processes. It's like calling your calculator a professor or a math genius. It's silly. Stop spreading this false hope of creating sentient, thinking beings. Even an infinite amount of monkeys smashing random keys on a keyboard would eventually pass the math tests. That's all that's happening here. Yes, it's cool, but don't oversell it.
    1