Everyone's losing their mind over the "new" Meta "Multimodal AI Glasses". I'm here to say it's already been out and round for 3 years.
We,
@LetsEnvision
have been building an AI-powered glasses for blind & low vision people since 2020.
Check this profile by
@BusinessInsider
@mrinininininini
Push back, ask them to speak in English. I’ve noticed these Hindi folks don’t respect language diversity all that much in the workplace 🙄
@ChrisJBakke
Lol. The management might be crying in joy thinking their little AI Chatbot idea is huge hit.
Little do they know a bunch of nerds are gaming it for free cars 😂
@dieworkwear
we're also extremely desensitised to seeing suffering that when we DO see the sweatshop that shein is, we simply say "that sucks" and go back to ordering that $10 sweater on limited-time discount smh.
@raiann77
@mrinininininini
I'd rather talk in a language that allows everyone to participate.
To most south indians like me, Hindi is a much more "foreign" than English.
@paraschopra
Tushita in Dharmshala. Tibetan Mahayana tradition.
When i went for the retreat, in 2012, they offered it for free to Indians because the Tibetans were grateful to India for giving them the space to build the monastery.
omg
@arrahman
dropped the most 🔥🔥 divine soundtrack in a loooooong time!
chamkila is such a beautiful album. You've got to start from the top and listen to it in exactly that order.
it's absolutely moving.
@ZachWeiner
@codinghorror
StumbleUpon. I yearn for the days when my web surfing wasn’t limited to just 4-5 websites. StumbleUpon was like hitchhiking across the web.
Being to tease out great music consistently from
@arrahman
is such a rare skillset. Only a dozen such ppl exist.
Imitiaz Ali is almost at the top of that list.
I was always skeptical about Chamkila but I'm very pleasantly surprised. I might not shut up about it for a while...
@MKBHD
that’s crazy! i can totally see a future where companies can use you likeness to generate a fake but very real-sounding video of you giving their product a very positive review. totally possible.
@ashleevance
your post hits on a few very valid points.
having spent 6 years building an ambitious AI + Accessibility startup in EU, i can attest to how inane the EU scene is. ofcourse, there are two sides to things. i will highlight the good at the end(cause it's obvious).
1. most EU VCs
@willdepue
He most likely has an agency or an intern do this. All for an exorbitant fee. Apparently it’s a thing tech founders do to “build an audience”.
The real Brett is somewhere in a coffee shop in Mission district answering his emails…
@paulg
This is assuming that SF has the best tech people. That might have been true 10 years ago but it's not true now, IMHO. Not being IN SF but having access to SF capital might mean a better use of $$.
@Suhail
Do you think that's just the twitter echo chamber talking this way? I think so.
I mean, according to folks here ChatGPT is stone-age technology whereas folks IRL haven't even heard of it yet.
Twitter is a bit ahead of the curve but not always right.
@anuatluru
goodreads has occupied this fantastic niche of too unprofitable to go after BUT low hanging enough to not fantasise building an alternative.
i suspect of all things on the web Goodreads will probably will remain untouched till the end of time 🤣
One of the best newsletters I've found in recent times -- Flow State. They send you really nice music to work to, every weekday morning, to your inbox. Links for Spotify, YT Music.
Definitely worth subscribing -
Apple is going all out with MLX.
a few days ago they rereleased MLX with Swift so you can run LLMs locally.
now they’re onto MLXServer so you can build APIs around them more easily.
solid TF/Pytorch competitor in the making.
Exciting new project: MLXServer
An easy way to get started with LLMs locally. HTTP endpoints for text generation, chat, converting models, and more.
Setup: pip install mlxserver
Docs:
Example:
@isabelunraveled
being really good readers of people. the women closest to me have a strong sense of intuition about the world and the characters in it.
the times i've listened to that intuition, i've benefitted immensely. saved me a few times from making horrible life choices.
@sidraqasim
it's infuriating and extremely depressing to think that i'm inhabiting the same piece of rock that such tragedies are happening.
that these heinous crimes are committed every single day, the folks committing them are going entirely unpunished.
Unfiltered:
Viduthalai is easily Vetrimaran’s most intense, political movie. My favorite movie of his.
The movie is largely dialogue-driven and that’s probably the best part about it. There were about 7-8 places that the audience hooted for the dialogues because they were on
@skalskip92
@Hine__Po
This used to be the hottest startup idea to fund in CV a 3-4 years ago. Now it’s a GitHub repo.
The pace at which things move is 🤯
a lot of folks are creeped out by it but this kind of tech is insanely great for accessibility.
In the short term, anyone with any level of motor impairment, can pair a
@neuralink
with an AI agent that can them get some amount of independence back.
On the day of its launch, we will introduce
@LetsEnvision
designed just for the 🍎 Vision Pro.
This app will be the first to fully utilize both the AI and accessibility features on the Vision Pro.
Over the past few months, we have collaborated closely with the Vision Pro and
@saffrontrail
Smart Watches are a big distraction. They shove metrics in your face and they're often incorrect.
Until this year, I leaned very heavily into the more data is good philosophy. It was overwhelming. In January, I stopped using the 🍎 Watch and I got into the best shape of my life
my favourite yearly ritual is to go to some coastal village in italy before peak season, bunk there for a full week in a b n b, and spend whole days at the beach alternating between swimming and reading.
i’m extremely grateful i can make space for it this year, in Sicily.
the
@abacaj
Wait. If your Rabbit calls are going to open ai how are they working out the unit economics of the device? I thought they would atleast have their own VLMs as a backup.
Are going to introduce a subscription later and currently burning VC $ to keep things afloat?
@OfficialLoganK
@OpenAI
We’re changing the way AI helps accessibility. We’re heavily using GPT-4 and GPT-4V on the Google glass hardware to help thousands of blind and low vision people to live more independently.
Check out this profile that
@BusinessInsider
made of us
Everyone's losing their mind over the "new" Meta "Multimodal AI Glasses". I'm here to say it's already been out and round for 3 years.
We,
@LetsEnvision
have been building an AI-powered glasses for blind & low vision people since 2020.
Check this profile by
@BusinessInsider
Excited to announce that
@LetsEnvision
has been working with the awesome folks at
@GoogleDeepMind
to build a realtime accessibility ally for blind people using an early access version of Gemini Flash 1.5. On Smartglasses.
we just got a shoutout at
#GoogleIO
keynote, more demos
@giuliofagiolini
@Spotify
you should just make a winamp-style player that lets people make open skins. spotify’s one chance to become part of tech lore forever.
@NickADobos
it’s way too eager to serve you clickbait. you like one racy picture the next refresh is 10x racier, one biden meme and my TL is teeming with alt-right shit.
whatever you like gets amplified into some extreme caricature, you’ve to fight hard to bring some normalcy to your feed.
GPT-V has the potential to revolutionize accessibility, specifically writing alt-texts for images.
Here's a simple chrome extension I built that uses
@OpenAI
's GPT-V to caption the main images on a webpage 🤯
Time to build that
@LetsEnvision
desktop extension
@samim
@jeremyphoward
Today I've picked up steam enough to confidently work on my own CV startup and the seed for it was sowed by fast.ai with Pt1. and Pt.2
cramped corner seat, snoring, spreading-their-legs co-passengers, 10 hr transatlantic flight with the worst food…
@arrahman
’s music is the only thing stopping me from the breaking the window and jumping off the plane.
divine, sonic blessing 🙏🏾
@abacaj
There needs to be a neat primer on using gen AI tools to create synthetic datasets for text/image tasks. Maybe there's one already and I'vent encountered it yet.
@OfirPress
@JiaweiLiu_
@steven_xia_
Most of these models have heavily gamed the leaderboard. I've been building stuff on top of Gemini Pro and it's definitely > 3.5 / < 4.
The OCR on this is really good.
@OfficialLoganK
Everybody should add alt-text to their images on social media. What with mutlimodal ChatGPT there's ZERO excuse to not make their work accessible to blind and low vision folks.
We built this nifty Alt Text Assistant that creates alt-texts for images.
@morganhousel
was my one of my favorite finds from the last couple of years. His writing had a noticeable impact about the way I think about my personal finance.
Lovely podcast by
@tferriss
👇
@fortelabs
This is like taking a sword to a pen fight. GPT Vision still hallucinates a lot for OCR. I’ve used Google Cloud OCR for this stuff and it works wonders.
there's a certain, sinister group of Silicon Valley insiders who absolutely look down upon every founder who hasn't
1/ done yc
2/ been to stanford/mit/princeton
3/ worked at FAANG
4/ subscribed to extreme libertarian views about biz and life
5/ b2b SaaS
🤷♂️
@mervenoyann
Here's what we use
@LetsEnvision
: we've built our own document layout detection engine(headlines, tables, figures, images) + proprietary OCR(Goog/Azure) pipeline.
we noticed it does so much better than using SOTA VLMs like GPT-4o/Claude 3.5 for this. Tesseract is a very poor
GPT-V has the potential to revolutionize accessibility, specifically writing alt-texts for images.
Here's a simple chrome extension I built that uses
@OpenAI
's GPT-V to caption the main images on a webpage 🤯
Time to build that
@LetsEnvision
desktop extension
🚀 Spent an unforgettable day with
@sama
and other devs from across EU.
Highlight was showcasing how tech like the
@LetsEnvision
Glasses, using GPT-4, can revolutionize independence for the blind and low-vision community.
Some exciting stuff in the works...
An assistant becomes truly conversational only when you can go below >200-millisecond latency. We managed to that today
@LetsEnvision
👀.
From query, to routing to response. Will work across both text, image & video inputs. More demos to follow!
@abacaj
I started building a complex RAG feature for my app a couple of months ago. The SOTA and frameworks have moved so fast even from back then they make my current setup look outdated.
Yet, it works well for my customers needs I feel no compulsion to change things. Fomo exists tho.
@sighswoon
100%
be curious not judgmental is more than just a cute aphorism, i believe. it’s the only sane way to navigate the fuzziest aspects of life.
@noampomsky
Lol.
In my grandma’s hometown in India there was a cow that was extremely charismatic. Everyone believed it was a good omen to pet it.
The owner of the cow turned around into a cult. Built a little shrine for it. Made enough from selling the milk and offerings at the shrine.
older i get the more i've started see the vanity metrics of life more clearly.
as i turn 32, perhaps my most important metric from here on would be vitality: to freeze my current energy levels for the next decade and if i'm lucky, beyond.
that alone would be my biggest success
@julien_c
I’m already using local LLMs on high end phones. One UX hurdle I’ve noticed is the size of the models putting off people. Plus after 5 minutes of usage phones tend to heat up quite a bit.
All of them solvable next year.
@sdw
Teenage me fell so hard in love with that UI I ended installing a Mac OS skin that rekt my poor windows vista PC.
Turns out skins on the internet were loaded with malware. Lost PC privileges for a few weeks over this but totally worth it.
The UI now is just bland.
Here's an early Thanksgiving gift from team
@LetsEnvision
! A new update is ready and now you can:
- Read text LIVE. In over 10 different languages.
- You can scan documents and read text in 63 languages.
- And a new UI!
Happy Holidays!