We are investors in Navana, which enables transactions to be closed via
vernacular voice. The company has been collecting voice samples since 2018
or so, with a focus on banking transactions and the like.They collected
various dialects and accents and colloquial phrasing.

There was a fantastic article on how empowered people feel when they can
complete transactions with ease. Isn’t this what all Indians deserve?

https://theprint-in.cdn.ampproject.org/c/s/theprint.in/ground-reports/voice-based-banking-for-low-literate-indians/2610584/?amp

IISc is also doing work on language sample collection.

Having a native language LLM is fantastic. That said, a lot can be done
just by unlocking the power of Gen AI for vernacular speakers.

And rock on, Ameya and Global Voices. Will make a donation.

Jo

On Fri, Jul 18, 2025 at 1:57 PM Ameya Nagarajan via Silklist <
[email protected]> wrote:

> Thank so much Venky!
>
> Re Tamil, there was this interesting report that came out recently about 
> content
> moderation in Tamil
> <https://cdt.org/insights/moderating-tamil-content-on-social-media/>,
> from the Center for Democracy and Technology.
>
>
> Cordially,
> Ameya Nagarajan
> (she/her)
>
> <http://www.linkedin.com/in/ameyann>
>
>
>
>
>
> On Fri, 18 Jul 2025 at 12:56, Venkatesh Hariharan via Silklist <
> [email protected]> wrote:
>
>> Radhika,
>>
>> Viswam.AI is working on Telugu LLMS while AI for TN
>> <https://aitamilnadu.org/> is working on Tamil LLMs. Viswam has worked
>> with around 45,000 volunteers to collect 1.5 million voice samples to train
>> their Automated Speech Recognition system and with colleges around
>> Hyderabad to collate text for LLMs. They are currently in the process of
>> training around 100,000 students in AI over a two month period. Around 20
>> percent of the time will be spent in collecting data for AI.
>>
>> I recently spoke to the Tamil LLM guys and the situation is challenging
>> but not insurmountable. Notes from my discussion below.
>>
>> Case Study: Tamil LLM
>> ● 70 million speakers
>> ● 15 trillion tokens needed to train Tamil LLM
>> ● Available Digital Tokens: 200 billion
>> ● Tamil Newspapers and Books: 2-3 trillion tokens
>> ● Need govt and community help to create more tokens
>> ● Compute is a challenge. Govt and industry help needed. Distributed
>> computing is an alternative model being explored
>>
>> Venky
>>
>> Ameya: I shall make a contribution to Global Voices over the weekend.
>>
>>
>>
>> On Fri, Jul 18, 2025 at 2:33 AM Yeddanapudi Radhika via Silklist <
>> [email protected]> wrote:
>>
>>> Hi Ameya et al! Just contributed in a small way to Global Voices. I was
>>> wondering if you know of any indigenous-led efforts in India to use AI to
>>> revitalize, restore, and preserve language. I read the story on the Global
>>> Voices site about Bultoo Radio (brilliant local name!) but that effort has
>>> the advantage of having many speakers. I'm thinking of efforts more like
>>> what Danielle Boyer has done:
>>>
>>> https://www.youtube.com/watch?v=KLRLyt5Ux7I
>>>
>>> Cheers.
>>> Radhika
>>>
>>> On Wed, Jul 16, 2025 at 9:33 PM Ameya Nagarajan via Silklist <
>>> [email protected]> wrote:
>>>
>>>> Dear Silklisters (S-listers?)
>>>>
>>>> Some of you know me, some of you don't. I'll admit I'm really more a
>>>> lurker here than a talker, but I am delurking to ask for help.
>>>>
>>>> I am the Managing Editor of Global Voices <http://www.globalvoices.org>,
>>>> a non profit that publishes news from the global south in many many
>>>> languages. Most of our writers and translators are volunteers, people who
>>>> really believe in how important it is to get this news out into the world,
>>>> to cross linguistic barriers and help people understand that really, most
>>>> of us have the same problems, and actually we need to band together 
>>>> against digital
>>>> authoritarianism
>>>> <https://advox.globalvoices.org/special/unfreedom-monitor/> more than
>>>> distrust regular people from one country over.
>>>>
>>>> GV and our partners work to defend the independence of the media
>>>> <http://advox.globalvoices.org>, the openness of the internet, and the
>>>> right of everyone, everywhere, to free expression. Our multilingual
>>>> newsroom publishes in 5 languages and stories are translated into over
>>>> 30 others <https://lingua.globalvoices.org/>. We work with Indigenous
>>>> activists
>>>> <https://rising.globalvoices.org/blog/2025/06/17/join-us-online-meeting-to-highlight-digital-initiatives-for-adivasi-languages-of-india/>
>>>> to decolonize the internet and digitize their languages, and raise
>>>> awareness of digital security. We are able to run really great projects,
>>>> like the climate justice fellowship
>>>> <https://globalvoices.org/special/global-voices-climate-justice-fellowship-2025/>
>>>> that pairs Sinophone journalists and global south journalists to write
>>>> investigative pieces about China's presence in other countries and the
>>>> climate impacts of it.
>>>>
>>>> We've been around for twenty years, and in 2025 we've been very badly
>>>> hit by Trump's funding cuts and their cascading consequences. We are a very
>>>> low-cost organization, with no office and few operating costs, and most of
>>>> our funding goes directly into staff pay.
>>>>
>>>> At a time when many for-profit media organizations are struggling,
>>>> Global Voices’ production and readership have been going up -- but since we
>>>> are committed to never paywalling and not showing ads, that doesn’t
>>>> translate into covering our running costs.
>>>>
>>>> The four years I have worked at Global Voices have transformed me, my
>>>> understanding of the world, and my faith in humanity. Yes, working in the
>>>> news in the past four years I have managed to have my faith in humanity
>>>> restored. That is the power of Global Voices. I can find another job, but I
>>>> don't know if we can ever build another Global Voices. Our mission is now
>>>> more urgent than ever. Please help us keep going! Donate if you can
>>>> <https://globalvoices.org/donate/>, and share in your networks.
>>>>
>>>> I apologize if this was not a good forum for this plea, but we're
>>>> desperate! Here are some recent stories that give you a sense of what we
>>>> do, pulled from the newsletter:
>>>>
>>>> Truth-to-power stories, like this one about an Australian journalist
>>>>> punished for her post on Gaza
>>>>> <https://globalvoices.us2.list-manage.com/track/click?u=9307106bf78cc186e924ad999&id=c2f49b4088&e=ff64988260>
>>>>> or this one about Turkey's political oppression
>>>>> <https://globalvoices.us2.list-manage.com/track/click?u=9307106bf78cc186e924ad999&id=643f270485&e=ff64988260>,
>>>>> remind me how much courage there is among journalists and activists. Other
>>>>> stories remind me about the complexity of the world beyond the headlines
>>>>> that are screamed over and over in my feed, like this one about 
>>>>> Uzbekistan's
>>>>> football team
>>>>> <https://globalvoices.us2.list-manage.com/track/click?u=9307106bf78cc186e924ad999&id=643f270485&e=ff64988260>,
>>>>> or this one about the de-colonialization of the official language in
>>>>> Niger
>>>>> <https://globalvoices.us2.list-manage.com/track/click?u=9307106bf78cc186e924ad999&id=55d6b513e3&e=ff64988260>.
>>>>> And then there are the articles that remind me of the joy in the world,
>>>>> like this one about International Reggae Day
>>>>> <https://globalvoices.us2.list-manage.com/track/click?u=9307106bf78cc186e924ad999&id=3dafc036e2&e=ff64988260>,
>>>>> this one about using K'iche’ Maya glyphs to reconnect art and language
>>>>> <https://globalvoices.us2.list-manage.com/track/click?u=9307106bf78cc186e924ad999&id=86ae3f66e8&e=ff64988260>,
>>>>> or the new special coverage about migration and belonging
>>>>> <https://globalvoices.us2.list-manage.com/track/click?u=9307106bf78cc186e924ad999&id=a5bd592612&e=ff64988260>
>>>>> .
>>>>
>>>>
>>>>
>>>> Cordially,
>>>> Ameya Nagarajan
>>>> (she/her)
>>>>
>>>> <http://www.linkedin.com/in/ameyann>
>>>>
>>>>
>>>>
>>>> --
>>>> Silklist mailing list
>>>> [email protected]
>>>> https://mailman.panix.com/listinfo.cgi/silklist
>>>>
>>>
>>>
>>> --
>>> *Translator/Owner*
>>> *AzulIndica Translations*
>>> *North Vancouver BC, Canada*
>>>
>>>
>>>
>>>
>>>
>>> --
>>> Silklist mailing list
>>> [email protected]
>>> https://mailman.panix.com/listinfo.cgi/silklist
>>>
>> --
>> Silklist mailing list
>> [email protected]
>> https://mailman.panix.com/listinfo.cgi/silklist
>>
> --
> Silklist mailing list
> [email protected]
> https://mailman.panix.com/listinfo.cgi/silklist
>
-- 
Silklist mailing list
[email protected]
https://mailman.panix.com/listinfo.cgi/silklist

Reply via email to