In this episode, Camb AI’s CTO and Co-founder, Akshat Prakash, talks about Camb AI’s revolutionizing language translation.
Ack Prakash is the CTO and Co-founder of Camb AI, a technology revolutionizing speech-to-speech translation with proprietary technologies and a focus on high-resource to low-resource languages. Here are a few of the topics we’ll discuss on this episode of Cache Flow:
- The importance of raising the bar for AI dubbing and translation by preserving contextual understanding and perfecting background scores, lip movements, and cross-lingual time syncs.
- How open source push, performance upkeep, and more activity of users will cause Pie Torch to win in the long term.
- How Apple’s lack of enterprise sales experience has hindered its potential to innovate in the cloud space.Why Camb AI focuses on zero-shot dubbing AI for converting high-resource languages to low-resource languages.
- Why Dubai is growing rapidly in AI and the Middle East because of funding and resources going towards it.
- 03:14 – A lower resource language, not necessarily isn’t spoken by fewer people like Malla is spoken by a lot of people. But it is true that perhaps it does not have as much digital data as something like Spanish would. Now, the problem that has existed is that more and more content as it has been developed in languages like English that are higher resourced to have a lot of digital data on the internet, speakers are not so fluent speakers of English. And, people who don’t speak the language at all often miss out on a lot of things that are developed in English. So, you know, high-quality education, health, sports, entertainment. But that’s just like kind of the tip of the iceberg.
- 04:54 – And with Camb AI, we’re hoping that with some of these foundational models that we’re creating that focus on this transfer of high resource to lower resource, we can provide essentially a backbone on which a lot of creators, enterprises, and general internet users can watch content or consume any kind of content in their preferred language and really, really democratize and build an internet that’s for everybody and not for somebody who speaks a particular language or, or doesn’t.
- 28:10 – It more depends on how much activity someone has. So for example, if Facebook is using, is the pioneer for Pie Torch, and they’re doing really kick-ass work in open source and doing a lot of good academic research, and they’re open sourcing that then that community will continue to grow. And I think this is a general principle for any AI system or framework over time.
- 35:52 – “I think all this stuff is going to continue to be more and more open source at the framework level, but I do think that the ways it’s going to be commercialized is that people will build products that have deep-rooted feature sets in AI capabilities. And then the other piece you touched on, which I love, is the models, the data. And, I think, there’ll be a concept in the future of like a mass. Like we have SaaS, we have PaaS, we have, you know, all this stuff. I think we’ll have MaaS models as a service”
- 1:04:49 – I think I try to remain very structured with my life because I know that if you do go too overboard, you’re not going to get work done anyway. You’re going to start becoming lazy. You never want to be in the place where you start disliking the love of your life kind of thing. Something that you put your blood, sweat, and tears into.
- 1:10:21 – I think what I’ve been reading and hearing more and more is that AI is not going to replace anybody, but people who know AI, understand it, or are able to work with it effectively are going to be the leaders of the future. And I just do encourage people to learn as much as they can be on podcasts like these and, and kind of other places where they can learn everything and anything about how this world is shaping out.
- 34:30 – Google was the original pioneer of this. I mean, tensor TensorFlow was like the OG framework for AI, and they were the creators, like I think Google has been at the lead of this charge for a long time.