Awesome progress at EX Synthetics

[Dollstudio]

Hi,

EX Doll's R&D with animatronics and robotics speeds up it's pace.

Compared with the protoype EX-ZERO there is quite some progress. It's no longer about getting things working, the focus is now on advanced areas like expression - and now even gestures. Currently EX is not limiting the work on an animatronics head, there is already a partially operational torso (I will post examples later).

Here is a short conversation with some body language:



Sandro

https://twitter.com/ExSynthetics/status ... 2603591680

[Dollpimp2018]

She looks great, they need to teach her English

[Dollstudio]

Hi,

yes, there is still much work left. However, the language with the most speakers is already covered - Chinese.

According to the Summer Institute for Linguistics (SIL) Ethnologue Survey (1999), this are the top languages by population:
(number of native speakers in parentheses)

1. Chinese (937,132,000)
2. Spanish (332,000,000)
3. English (322,000,000)
4. Bengali (189,000,000)
5. Hindi/Urdu (182,000,000)
6. Arabic (174,950,000)
7. Portuguese (170,000,000)
8. Russian (170,000,000)
9. Japanese (125,000,000)
10. German (98,000,000)

Dr. Bernard Comrie’s article for the Encarta Encyclopedia (1998) gives slightly different numbers (number of native speakers in parentheses):

1. Mandarin Chinese (836 million)
2. Hindi (333 million)
3. Spanish (332 million)
4. English (322 million)
5. Bengali (189 million)
6. Arabic (186 million)
7. Russian (170 million)
8. Portuguese (170 million)
9. Japanese (125 million)
10. German (98 million)

Sandro

[MannyCan]

She looks great, they need to teach her English

Or do you need to learn Chinese?

[Dollstudio]

Or do you need to learn Chinese?

…or even better, maybe these synthetic girls can TEACH us Chinese at some point in the future?

Sandro

[AhMyDolless]

Or do you need to learn Chinese?

…or even better, maybe these synthetic girls can TEACH us Chinese at some point in the future?

Sandro

That's actually a really good idea. A foreign language speaking buddy doll. Especially for harder languages like a Mandarin/cantonese and hardest of all Icelandic.

[Willendorph]

She is really looking good. I think the head and neck movement looks a lot smoother in that video than in the previous ones, which is really nice as I thought that was the biggest area that needed work.

With regards to languages, they're going to go where the market is (for example if loads of people speak Hindi, but they don't buy dolls then it'd be pointless). I'm sure English will be next, but In the perfect world, I'd really want her to be bilingual and have Japanese as well.

Regardless of which way they go next, I know I'll be watching with interest.

[NeedWaifu]

Sandro, do you know. Are they focus on random generated talking or pre-recorded dialogs? An which is cheaper technology? I think the latter would hit the market a lot quicker?

[Dollstudio]

[…] Are they focus on random generated talking or pre-recorded dialogs?

For me the key question for anything seriously robotic is if the dialogs respectively movements are AI or scripted. If I am not totally mistaken, everything we get to see is still scripted.

The other question would be if there is true speech recognition and speech synthesis. For that the tech exists, but I haven't seen it yet embedded in a humanoid head. Processing usually takes place elsewhere, either wireless on a local server or even remotely somewhere in the 'cloud'. Since all of this is still R&D, I guess most of it happens locally.

Pre-recorded samples from organic humans would severely limit the scope of a serious robot. Existing products like Z-Onedoll and some TPE dolls have options for this. I am pretty sure that this is not the level EX is working on. EX is way too ambitious to just reach for the lowest hanging fruits.

For the video above, I was curious what this dialogue is about. I have learned some Chinese back at the university, but this was a long time ago, so I understand basically the same as everybody else here - gibberish.

So tech to the rescue… Extracted the audio from the video and fed it to the IBM Watson Speech to Text service. Watson gives me:

我想买一个新的。
唱一万多拉。
你有网友啊。
他还没。
那天去抢去参军。

(the voice model for Mandarin does not detect multiple speakers, so it's all in one line)

Now ask Google Translate:

I want to buy a new one.
Sing more than 10,000.
You have friends.
He hasn't.
I went to get involved in the army that day.

Anyone understand this? I don't.

I think that's the state of the (available) tech. And I guess that answers the question how much of these sample videos can be not scripted…

Sandro

[nukeno]

The other question would be if there is true speech recognition and speech synthesis. For that the tech exists, but I haven't seen it yet embedded in a humanoid head. Processing usually takes place elsewhere, either wireless on a local server or even remotely somewhere in the 'cloud'. Since all of this is still R&D, I guess most of it happens locally.

I would love to see that technology one day.
However, it will be most likely cloud-based, since you need AI learning and big data sets to get away from scripted dialogs.
You cannot do that locally. And that's what freaks me out the most. On one hand, the internet is mostly for porn already anyway and telephone sex or chatting with cam girls is not that different.

However, if you have one big centralized AI, something like Amazons Alexa, controlling all the sex-robots, hearing every word and "moan" you make... scary!
And on the other hand very different from having telephone or webcam interaction with "just" one other person over a more or less secure line or server.

I now, it won't bother many people, like the privacy invasion via smartphone or Alexa does not bother many people today.
However as long as I cannot have truly private/offline moments with my "robot", there is no way I would do it, as much as I would love to.

[nukeno]

Some people believe that there is a god (or Santa) who knows everything, sees everything, hears everything.
So some may teach their children things like: "masturbation makes you blind (or what ever)" and that you'll go to hell if you do it!
However, these consequences only happen in the afterlife (or once a year at Xmas).
With an all knowing, seeing, hearing AI, we will see consequences in our daily lives, and it will be a living hell.
The end is nigh!

[Dollstudio]

Hi,

take 2…



Same dialogue, I think. So yes, it's scripted.

Sandro

https://twitter.com/ExSynthetics/status ... 2472379393

[Dollstudio]

Hi,

some more dialogue:



Sandro

https://twitter.com/ExSynthetics/status ... 3378453504

[Altair]

For those curious about what the dialogue was about.

In the first two videos, the topic was regarding the iPhone.

Doll: 哥，我想买个新的iPhone.. 差一万多嘛
Big bro, I'd like a new iPhone... but I'm lacking 10 grand!
Man: (some question)
-was talking very fast so I couldn't tell what he was asking
Doll: 有啊有啊
Yes, I do have that
Man: (some question)
-same as before, he's talking very fast as most native speakers do
Doll: 还没嘛还没嘛
Not yet
Man: 你赶紧去抢啊，赶紧去，赶紧去
Then you better get going and grab one of those!


The third video seems to be comedic in nature

Man: 儿子啊，你最近成绩很稳啊~ 59分
Son, your result's really steady isn't it? You scored 59 marks!
Man: 打游戏打的吧
Is that the result of playing video games?
Doll: 哈？游戏什么的我从来都不玩啊
Wut? I never played those games before
Man: 那就是学校学到你的坏了吗？
-not sure if I heard that correctly, I think it's about picking up bad habits in school
Doll: 那当然，随便考
Of course, I don't take exams seriously.

The following dialogue was about Chinese idioms
-Made of 4 syllables
-Usually the 2nd and the 4th syllable would rhyme
-Part of the exam syllabus in China
-The interaction between the man and doll is sort of a game/test played by 2 people to see if one could remember the idiom by coming up with the last two syllables based on the first two provided.
-The man provides the first two syllables while the doll completes it.
Man: 雕虫 Doll: 小技
Insignificant skill
Man: 忘恩 Doll: 负义
Ungrateful
Man: 低声 Doll: 下气
Submissive
Man: 大吉 Doll: 大利
Prosperity
Man: 今晚 Doll: 吃鸡
Not an actual idiom. Literally meaning "I'm having chicken for dinner"

At this point, the doll shows a shocked expression after 'realizing something'...
-This is because the last 2 'idioms' were a trick question.
-大吉大利 on its own is a legit idiom. But when combined with 今晚吃鸡...
-It becomes "大吉大利，今晚吃鸡"
-This is a term from the popular game PUBG, used by chinese players to refer to the game and/or identify fellow players.

Earlier, 'Big brother' questioned the doll if she was playing video games, which she denied.
However when she managed to come up with the term "大吉大利，今晚吃鸡" which only PUBG players would know, she 'realized' that her cover was blown.

[Dollstudio]

@Altair: Awesome, many thanks for the translations!

Sandro