Відео

That way you can keep your typical custom instructions

Why?

@@peterwilkinson1975 Because AI is disengaging, bland, condescending and kinda painful to watch, while random bugs are interesting, have a purpose and a struggle in life, like the rest of us.

@@ZappyOh how much of it is the fear of change? Or the world we live in be different? Eventually ai will be smarter than us, so we won’t be able to compete when it comes to creativity or anything. In the same ways monkeys can’t compete with us in creativity. Reason why is ai lives on a different time scale. Millions of years can go by for ai in what are hours for us. You might think that will never happen. It’s already been solved, in that we know how to achieve it. It’s just a matter of having the primitive version design the chips it runs on (already happening). Then the learning component teach itself how to learn better. Learn how to learn, then learn how to improve the way it learns. Then learn how to improve that. This isn’t theoretical we have proven this is possible, it’s just a matter of a couple years. Building better hardware ect. You might be like well this is some random youtube commenter, I’ll choose to ignore them. What I’m saying is true, I don’t have to convince you because reality will do that for me.

@@peterwilkinson1975 Sure, straw-man me, and everyone not loving AI. Here is an idea: Ask AI to analyse your love of it, and see what it comes up with ... then ask a real psychologist to do the same, and compare notes. But please, don't psykoanalyse your commenters. It is rude, and pretty cringe when you obviously don't know what you are talking about.

@@ZappyOh see your looking to discredit me to preserve your worldview. Reason why I am making the assumption about your motivations is I see literally hundreds of people doing the same thing. Not just about ai you see the same thing with evs ect. New styles of music, when the Industrial Revolution happened ect. It’s the natural response to change. People think the world is normal (not going to change). Then fight to preserve the normal.

Ha! Oh, damn. Nailed it.

Tokens are free if you run the models at home. We need ways to condense the large models down to something that can run on hardware we already have.

Yeah def. Cost engineering is a big part of it atm, and probably will be for awhile. That said, 4o-mini is quite capable for some things and incredibly cheap. Have you played around with fine tuning that model?

Even easier, just add it to memory

He would never do that. She's not a whore 😏

Yep, but whatever I guess.

I skipped most of his rant lol.

yeah but only because the current rate limits. that exact state plus at least 3 hours changes a lot. It's the start of a new era. A complete change on how we interact with tech even though we justt had that with GPT-3.5. Now imagine you talk to the 100x times larger Osiris in the most natural non prompt engineered way, give it all contexts by talking to it for 5 minutes about an idea or something. an then it goes "alright. le'mme think about that" and then it goes thinking until it has a good solution. No wonder they call it a new era themselfes

Memory works.

Yes. It definitely does. Context is basically what the “this thing is smart” reaction comes from. If you give it no context, it’ll seem “dumb or robotic”.

@Toxicflu Giving it context will make it feel/act smarter than without any context or information - so giving at all the context makes it 'smarter 'in that particular use case...

Besides, that's not the point of the advanced voice update... It's the voice, not the intelligence.

LOL. She asked me a verrry open ended first question. I didn’t want to disappoint her 😂

​@@MikeRobertsmrspynot wanting to disappoint her is the reason I'll probably never interact with AI if it ever becomes sentient. I'm not sure there's any value I can ever provide something that is so intelligent, that I'm not simply wasting it's time 😄

It's a good idea. I'm going to test this, for sure.

Right. The way you expect it to work is exactly the way I thought it would. But, it does not. When you do that it silently reverts to basic voice mode. Which definitely fooled me given that I’d just been welcomed to the new Advanced Voice mode in app. Yeah - got excited to test it and forgot to turn off notifications.

Nope @MartinKoss . If you try to switch any text-based chat to an AVM chat, it says it 'cannot use the data' and asks you to start a conversation from scratch. So there is no easy way to give it advance context or info (apart from custom instructions if that gives you enough characters).

Oh, Interesting idea. I’ll try that. I was pretty surprised that I couldn’t just give it some context via the chat, so I thought it must be starting completely from scratch. But, I did notice it can access its memory of me.

That idea could be limited by fact custom instructions are limited to just 1500 characters of information about you.

​@@RJMCTV yikes that's not much

This works but is very limited context. I have used this to get around the issue but it’s a pain.

@@RJMCTV 1500 for "things about you" and 1500 for "how you want Chat GPT to react. You can use both.

FWIW, it feels quite a bit less smart than basic mode. I suspect this might be 4o mini and basic mode is 4o. I don’t have any evidence for this either (despite a quick search)- it’s just a feel. I asked gpt advance voice mode, and she insists she’s GPT4 even when tell her that’s definitely not the case. 😂

Yeah. You can’t - when you do, it switches back to basic mode.

Nope. No vpn.

I definitely will.

you also were able to pick up the core intention of his video 👍 out of curiosity, i browsed your channel and found out we have some common interests (3d, synthesizers, "the 8-bit guy", "Xiaoma nyc" , etc) Have a good day! 🙂

Glad you got it, thanks! It's definitely about how to use it. I feel like whether you're doing a podcast or not -- this could be a really useful way to "pull" content out that's sort of stuck in our "wisdom and experience". Writer's block? Don't have ChatGPT write it -- have ChatGPT ask you questions.

@@MikeRobertsmrspy A perfect use for it 👌

It’s 4o mode in ChatGPT which is using the 4 voice, yeah.

@@MikeRobertsmrspy it is NOT using the voice mode that was demoed by OpenAI...

@@danielkahbe964 Yes, that's correct. This is the latest version of ChatGPT in 4o language mode, which uses the 4 voice.

It does have the 128K text model which allows it to carry this conversation for 45 minutes.

No. It is GPT-4o using the old voice mode. The new voice mode will roll out in a few weeks.

Yeah, that’s correct. It is 4o - mode, but the voice mode will improve. I’ll test it out when it does.

The VTT and TTV model are layered on top of GPT4o , like the old Gpt4 the new Native MultiModal one with less than 250ms latency and the Vision + Emotion capability is not yet out for public except the Red Teams

@@lawyermahaprasad ik

they haven’t rolled out the voice model

Yeah, should get better when the 4o voice is released. This is 4o mode w 4 voice.

LOL. Love you kid!

You'll get used to it, way more than boomers lol

@@markmuller7962 i think Mike is 40yo, so he's a generation ahead (Gen X)

​@@webgpu I wasn't talking of Mike Edit: And btw 40yo is Millenials not genX

@@webgpu he’s 46

OMG yes. I don't think I've got time, though. I feel like it's a pretty obvious need, and I'm gonna bet it gets filled.