Hume AI (quick review)
HTML-код
- Опубликовано: 10 апр 2024
- I had heard about this company in 2022 but it was mostly data demos and papers. This new demo is amazing and the application for this kind of tech gives me so many ideas.
I did this screen capture a few minutes after first testing it so it's a bit wacky. For timing I did cut out the filler words and silence. The responses from Hume were still very fast.
So excited to see where this tech goes. 
Наука
This is awesome thanks for sharing
So cool right? There’s even facial expression detection. It’s wild.
Brilliant demo! Hume AI's emotional intelligence and natural conversation abilities are mindblowing. Your video highlights this emerging tech perfectly. More like this please!
Right? I feel like I barely scratched the surface. I’ll definitely share more. Thanks.
This AI sounded really cheesy and too verbose. Yet, it's still awesome. I hope they keep improving that. Seems kinda cool, for some reason doesn't give much uncanny valley.
Yeah his personality was a bit cheesy but I still enjoyed it. I’m sure it’ll only get better, perhaps a non cheesy option. :)
The verbosity didn't bother me because I admit that I am a bit on the verbose side myself. Definitely sounded friendly and cheesy and over the top which is definitely what I'm used to as an ADHD. I wonder if we could program an ADHD ai and then test to see whether or not ADHDers have an easier or more difficult time following it.
@@Dragoon91786 oh that’s an interesting take. Having personas with specific characteristics and even customizing that.
@@RetropunkAI There's definitely high risk of our languages (particularly, those of less powerful social groups) destroyed as computational linguist and NYU professor Michael Gasser spoke of quite at length last year¹.
In response to my question about if open-source, community controlled language models could help reduce or eliminate the numerous problems he illustrated throughout his own lecture (And from my perspective they are definitely worth being heard by any and all people involved in AI, which is probably why there was at least one Google employee there 🤣), Gasser said that he thought that it might be possible. He said the biggest problem is the power is so centralized and focused into the hands of people who are diametrically opposed to the welfare of the minorities of the majority of people on this planet (paraphrase on my part). He was skeptical, but he said if any good was to come out of these systems and not totally fuck up our planet and cultures for the worse, it would only be through the most democratic, universal, distributed methods whereby every community can take advantage of and insulate themselves from these risks. However, at present at least for many of these models, they don't solve the giant environmental costs that, like cryptocurrency, are produced in order to make them a reality.
Though, I think those wafer sizes chips probably could help some.
I think open models, open hardware and software, and truly public semiconductor fabs is really the only hope we have for the models not to be used against the majority of people in ways even those designers and implementers are likely to realize.
We want AI to be a force for social progress, not collective suffering. 😅
BTW, one concept that came to me was based upon the deaf students in a South American nation (I don't remember the name off the top of my head) whose indigenous) peoples were sent to a single school together (I think this might have been Venezuela) and over the course of several school groups, the children created their own sign language and taught it to their peer-evolving it rapidly over time, and giving linguists a far greater understanding for the evolution and development of language.
My thoughts were that "persona" could be useful for the diversity of human consciousness and the numerous ways through which human beings operate-cognitive diversity.
Not every speaker is going to be adept at communicating with every single spectrum of people. Like, if I take Star Trek as a metaphor, not everyone can think like a Vulcan or Romulan or new antiauthorian Borg "drone". Or the extra galactic species introduced last season in Discovery. So, how could you possibly hope to communicate effectively with, Grok, comprehend that person or persons? Wouldn't you need models that more effectively model the broadest range of the human experience? It would need to be able to model ADHDers, Autistic Spectrum folks, people growing up in multigenerational C-PTSD infested communities and I those people for whom these aspects are totally foreign, and so much more. Cultures that focus on the importance of family vs work, those that seek for greater meaning, let alone the wealth of human knowledge, writings, thoughts, etc.
And it certainly isn't going to help it to only meet some cognitive normative baseline like it's a European ancestry, upper middle class (white) cis male or female of a particular generation. Not will it likely help for it to be the scrapings of the worst parts of the internet.
I imagine models that could help "translate" or help provide learning opportunities for free to people (say with different forms of cognition) provided the free and universal aspects of the models and methods are maintained.
Would be great to have something that can translate my ADHD verbosity into something a super concise, no patience person can readily understand and follow. Maybe, helping translate stuff for us? It's got immense risks that would need to be addressed, and a pox on the sorts of rentier finance capitalists that want to bring about our worst nightmares, but IF we SERIOUSLY take into consideration the potential risks and collectively work with and through our respective communities in truly free, open, and collaborative ways, then this AI stuff might yet prove of social consequential value and not instead as a means by which oligarchs can steal/concentrate more wealth into their hands until *_Max Headroom_* is no longet that quirky dystopian Sci-Fi comedy series that got used for a famous hacking, and instead our waking reality.
1. Assuming the link goes through, his lecture/QA is available Via the People's Forum, a New York City staple and where you can get excellent left of center academic and social commentary from activists in the field. I use it to round myself out; the courses by fellow NYU professor and economist, David Harvey on Marx's *_Capital_* has been particularly insightful.² peoplesforum.org/events/language-technology-and-the-digital-divide-science-against-capitalism/
2. peoplesforum.org/capitaldavidharvey/
Despite what people might claim, in general they don't come across as tankies, and it's a useful resource for understanding a broader range of people (certainly important if someone wishes to effectively model a broad range of the human condition through different AI).
@@RetropunkAI posted a reply, but wasn't sure if the links I shared to related lecture topics (at least, for my ADHD brain) caused it to glitch. Please, let me know if it doesn't post so that I can try again without the links. :3