This reminds me of the project at IIT where they were building a voice
biometric lock. Someone watched Alibaba and the forty thieves and decided
that voice biometrics were the future. It would be odd, walking down the
hallway in IIT Madras and suddenly hearing a poor student intern try
variations of "Open Sesame" to regression test the lock. An army of them
were hired and even encouraged to do mimicry. It always lightened the mood
passing by that corridor.

In an environment where you might be working along with others, suddenly
uttering some secret phrase might not be the most conducive, isn't it? Also
if you have roughly five minutes of someone's audio, that is just about
enough to train AI to clone their voice.

On Thu, Nov 5, 2026 at 11:10 AM Advaith Mohan via Silklist <
[email protected]> wrote:

> I was imagining an experience, where instead of setting a password, you
> just speak for a minute, or till whatever time is needed to establish a
> unique voice + pattern print. But yes, that too will increasingly be
> AI-fied.
>
> Regards,
> Advaith
>
>
> On Thu, 5 Nov 2026 at 09:41, Udhay Shankar N <[email protected]> wrote:
>
>> On Thu, Nov 5, 2026 at 9:14 AM Advaith Mohan via Silklist <
>> [email protected]> wrote:
>>
>> Claude's recent text watermarking algorithm
>>> <https://www.anthropic.com/news/claude-text-watermark> made me think
>>> that a system analyzing our writing and/or speaking patterns to create a
>>> strong user ID, would be a good backup to the traditional password.
>>>
>>
>> It is not clear that the majority of people actually write their own
>> (e.g) email, instead outsourcing the process to some LLM.
>>
>> Udhay
>>
> --
> 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