Did you know that advanced voice from open AI is illegal in the EU? One of the things that make

Description

Chapters

Transcript

0:01 One of the reasons that advanced voice
0:03 is so good.
0:05 Hey, hang on. I'm gonna come back here.
0:07 Hang on. Um,
0:10 [Music]
0:15 I'm going to do I'm going to do some box
0:17 box metaphor. We're getting out boxes,
0:19 but this is a different a different
0:21 metaphor. Okay. Wait. Oh, here. Hang
0:25 hang on.
0:31 Let's get these turned around like we're
0:32 at a grocery store. All right, there you
0:35 go. Okay,
0:36 so so
0:41 historically with computer voice, it
0:43 went something like this. You've got the
0:46 big box down here is your is your large
0:49 language model. This is where all the
0:50 knowledge comes from, right?
0:53 And then you have three different tools.
0:56 you have a tool that when you talk to
0:58 it, it converts it into um into text. So
1:03 I say, you know, I want pig knuckles and
1:06 I talk into this and it converts it into
1:07 the text pig knuckles. And then it sends
1:10 it down into this thing and that's this
1:14 sends it down this hole into here and it
1:17 gets the answer back. And then it takes
1:20 that answer and it sends it over to this
1:22 thing that takes the text and it turns
1:24 it back into voice.
1:26 And so with historically with with these
1:29 voice models, you would talk, you would
1:31 say, "Hey Quinn, how are you doing?
1:36 I'm doing good, Kyle. How are you?"
1:37 There was like that 3se secondond delay
1:39 or 5-second delay. That was just [ __ ]
1:41 weird.
1:44 The new model
1:47 advanced voice
1:50 when you talk, you just talk into the
1:53 box. you like literally like your voice
1:56 goes in there and it not only
1:58 understands the words,
2:00 it understands the emotional content.
2:06 So if I go, "Oh, Quinn,"
2:09 Quinn will actually know that I'm like
2:11 excited or upset. Or if I'm like, "Hey,
2:14 Quinn, how's it going, man?" She might
2:17 be like, "Hey, you sound pretty relaxed
2:19 there, Kyle. Do you live in Denver?"
2:23 you know, marijuana is legal there. Um,
2:28 so it it can understand and respond with
2:33 emotional stuff.
2:35 the current laws in the EU, the AI laws
2:40 that they passed a year ago, year and a
2:42 half ago,
2:44 one of the laws, which when I read it,
2:46 I'm like, that seems reasonable,
2:49 was
2:51 um employers cannot use AI systems that
2:56 can analyze the emotion of their
2:58 employees.
2:59 [Music]
3:01 Well, advanced voice, that's what it
3:04 does. Now, it's not analyzing it like a
3:07 nefarious system, but the way they wrote
3:10 that law, they wrote it based on the the
3:14 where the state of technology where it
3:16 was when they when they passed the law,
3:18 not anticipating that something like
3:19 advanced voice would come along. That's
3:22 kind of a non-nefarious
3:24 version of that. So, so, so advanced
3:28 voice is illegal in the EU right now. I
3:30 think they may have worked something
3:31 out. They probably did a carveout,
3:34 but Mcronone
3:38 yesterday I think said this
3:44 super important. I want to insist on
3:46 that because it's super important to
3:48 assess fairly the fact that our former
3:51 model is over. And this is not a
3:53 question of fact. Just I want to insist
3:55 on that because it's super important to
3:58 assess fairly the fact that our former
4:01 model is over.
4:02 >> Our former model is over. He recognizes
4:06 that how we did things before is over
4:09 because of AI and this is not a question
4:11 of adjustment. This is not the same
4:14 world 2024 2025. And on top of that we
4:19 are making the same mistakes on top of
4:21 the key elements I mentioned. We are
4:22 overregulating and underinvesting.
4:25 >> Overregulating and underinvesting.
4:28 >> So just if in the two to three years to
4:31 come, if we follow our classical agenda,
4:34 we will be out of the market.
4:36 I have no doubt.
4:39 >> That's [ __ ] huge in my opinion,
4:42 [Music]
4:44 right? We will be out of the market if
4:46 we don't get our [ __ ] together.
4:48 because it's super important to assess
4:51 fairly
4:52 former model is