chris
@hingeloss
Followers
2,591
Following
1,064
Media
390
Statuses
3,669
optimism of the will, pessimism of the intellect.
NYC
Joined December 2016
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Tim Walz
• 1144425 Tweets
Your Season By NuNew
• 341306 Tweets
Shapiro
• 287320 Tweets
ホロライブ
• 262030 Tweets
湊あくあ
• 225079 Tweets
Minneapolis
• 177789 Tweets
Australia
• 121421 Tweets
LISA FEAT ROSALÍA
• 121007 Tweets
National Guard
• 107651 Tweets
#LISAxNEWWOMAN
• 104770 Tweets
Olmo
• 95121 Tweets
George Floyd
• 74101 Tweets
#西園寺さんは家事をしない
• 68820 Tweets
Aqua
• 63425 Tweets
あくたんの卒業
• 60213 Tweets
Deco
• 50882 Tweets
#WonderFOURyearswithTREASURE
• 48479 Tweets
#빛나는_트레저_4번째_생일축하해
• 48106 Tweets
卒業発表
• 33058 Tweets
Serbia
• 30108 Tweets
方向性の違い
• 29570 Tweets
ギラホス
• 22383 Tweets
アドゥワ
• 20041 Tweets
ホロメン
• 17764 Tweets
対象2作品
• 17283 Tweets
Jokic
• 15041 Tweets
Pinned Tweet
Presenting: the world's fastest AI voice chat - 500ms latency, running locally, 2x faster than anyone else.
How is this possible? 👇
97
226
2K
Food $200
Data $150
Rent $800
Substack newsletters $3600
Utility $150
Somebody who is good at the economy please help me budget this. my family is dying
20
20
584
@JayaGup10
gonna start an "AI daycare" where sf girls drop off their bf's: they can all talk to each other and get the AI out of their system. it's free, we just take 5% of any company they start
7
11
306
@Nexuist
one chart you want number go up, one chart you want number go down. which way, western man?
1
8
293
This demo uses Gazelle, the world's first public LLM with direct audio input. By skipping transcription decoding, we save time and can operate directly on speech - with inflection, tone, emotion, etc.
🔊Excited to ship the latest iteration of Gazelle - we now process and reply to spoken commands, with <150ms latency. This model shows strong reasoning capability and effectively generalizes to new tasks.
Want to try the demo? Thread!
4
20
173
7
22
236
🔊Excited to ship the latest iteration of Gazelle - we now process and reply to spoken commands, with <150ms latency. This model shows strong reasoning capability and effectively generalizes to new tasks.
Want to try the demo? Thread!
4
20
173
You can tell this is a real VC account because of how they have no conviction and change their mind on the herd of "the best firms in the valley"
update on harvey:
they just announced $100m in new funding from insiders (including all the best firms in the valley).
that’s a much more powerful signal than anything i/the harvey competitors who have run with my prior tweet have to say.
35
6
305
3
1
177
low serotonin week. think i'm going to buy something absurdly unnecessary from ssense
3
4
139
I've built and worked on trillion scale infra. The number one performance lesson is always to _reduce variance_ first -- simpler architectures with fewer components will always win.
The ASR-LLM-TTS cascaded systems will never be viable.
5
9
132
one time I interviewed a Twitter ML influencer and they bombed a leetcode easy. for a long time, my takeaway was 'those who can do, those who can't poast' but perhaps it should have been 'publicity is the best way to get roles you're unqualified for'
8
2
124
Very excited to see OpenAI launch voice-to-voice: humanlike conversation is one of the most important problems of our time, and simpler architectures win.
Nowadays, few people publish how cutting-edge models work. Here’s my explanation for the end-to-end approach, some thoughts
6
9
121
Instruct tuned 405B sets SOTA for MMLU-Pro (!) Still seems to be a bit behind 3.5 Sonnet on the other hard evals, but very much in the same ballpark. Can't wait to vibe-check it.
Also notably -- new license () removes the prohibition on using Llama 3 to
Compared leaked Llama 3.1 benchmarks with other leading models, very excited for the release!
We can tier out models by price / 1M output tokens.
O($0.10): 4o-mini and <10B param models. I think 4o-mini will still be best but a strong local 8B will unlock lots of applications.
1
4
30
4
16
121
👀👀found a hidden Microsoft page with a new TTS model, VALL-E 2, which claims to 'achieve human parity for the first time.'
AFAICT the paper is not out but description/snippets released 2024-05-20.
6
8
116
In SF next week - I'm told nobody drinks anymore and sleeps early, so what's there to do? Cold plunge?
28
0
115
quick Llama-3 8B throughput testing with different GPU's on
@modal_labs
- cheaper than dedicated providers, YMMV, script/method in replies
TLDR: H100 is much more cost efficient than A100. fp8 kv-cache is ~5% more throughput.
11
13
105
After that, it's bog standard optimization work - our implementation is very close to if not SOTA for multimodal LLM inference, and close to theoretical maximums.
With an H100, we expect this experience to be <300 ms - below human reaction time!
3
4
95
yeah I'm working on the frontier of AI (googling pytorch errors that only me and one FB engineer have run into)
1
5
84
Obviously this particular model is undertrained and there's a lot of room for improvement, but I'm very confident this is the future of voice AI.
What would you do with truly real-time and empathetic chat?
11
1
73
I work in big tech. A name you have heard of and probably used before.
Instead of performance reviews, my boss handed me a business card with threes shapes. I'm now on an island with the other engineers and only one of us can make staff. Wish me luck!
2
4
72
LLMs are officially in the web3 era
- undergrads prominently flaunting Stanford creds
- non technical guys writing white papers brought on to "promote and market"
- copy pasted code and weights
Only thing missing was exit liquidity
So sad to hear the news ()😰. The conclusion of our investigation:
1. Llama3-V can be run using MiniCPM-Llama3-V 2.5's code and config.json after changing param names
2. It behaves similarly to MiniCPM-Llama3-V 2.5 in unrevealed experimental features
13
73
500
3
0
69
🔊 I'm releasing a research preview of Gazelle, a unified speech-language model. It's not actually good (yet!) but already does tasks no other model can do - including even Gemini, Claude or GPT4. Thread!
6
6
62
@soumithchintala
@SkyLi0n
a googler's response (paraphrased): "nowadays we just copy-paste code. no need for unit tests, python typing, or readability review"
feels short term positive long term negative
1
1
56
I trained a joint speech-language model that you can *talk* to - for less than the price of a Chipotle bowl. Why I think this is the future of conversational AI and where we go from here: 🧵
4
7
53
@aquariusacquah
Clever hack to improve margins: no credit cards accepted, only cash and P2P payment apps!
1
1
53
I read the Thiel social network pitch deck so you don’t have to. Prediction: the
@deadspin
alumni blog this weekend will get more engagement than Column ever will
3
5
49
once I was a data scientist and got good at it and became a PM (playing google docs)
then I was an ML engineer and got good at it and became a tech lead (playing google docs)
now I am a founder and am getting better at it, and once again, I just play google docs (and linkedin)
2
0
48
@aquariusacquah
The “you don’t pay the plumber to hit the pipes, you pay them to know which pipes to hit” theory is gonna get tested real fast
0
1
45
Meta trained a E2E speech experience with Llama 3.1 - pretty cool! This should equal real-time speech response.
Audio encoder + adapter + LLM = audio in, text out
Custom TTS model uses LLM embeddings to condition output - IMO elegant to stay in latent space and avoid phonemes.
4
5
46
Most of the magic is from custom inference code and orchestration but you can play with the base model today on !
Presenting: the world's fastest AI voice chat - 500ms latency, running locally, 2x faster than anyone else.
How is this possible? 👇
97
226
2K
2
2
44
haven't coded in almost 2 weeks, already feel the autism fading away and the social skills returning 🥲🥲🥲
5
0
42
@wyatt_privilege
@barry
Emperor's new clothes, but if any criticism of the royal family was a heavily enforced felony
0
1
38
cannot believe the LA Rams quarterback won an NFL game despite spending the preseason studying Machine Learning
0
0
38
@felix_red_panda
@jordi_cor
@nonagonono
Imo most pragmatic path is to skip whisper/transcription entirely
3
2
37
0
1
33
I got one (1) compliment recently so I'm set for the next couple months
2
0
32
@sir_gee_ohhhhh
@grlalx
This would be excusable if their target market wasn't "people who need prescription glasses"
1
1
32
standing offer for Flatiron-ish AI folks: Caffe Panna and Gramercy Park on me :)
Today I’m thrilled to announce
@Lux_Capital
's NYC AI Directory & NYC AI Map - 2 resources for the burgeoning AI talent ecosystem
READ MORE👇
NYC AI Directory:
NYC AI Map:
30
53
356
6
0
31
Compared leaked Llama 3.1 benchmarks with other leading models, very excited for the release!
We can tier out models by price / 1M output tokens.
O($0.10): 4o-mini and <10B param models. I think 4o-mini will still be best but a strong local 8B will unlock lots of applications.
1
4
30
“I spent five years studying economics because it was the easiest major that pleased my Asian parents” -
@FunnyAsianDude
and also me
1
3
28
The best part of doing a startup is getting to choose the right thing over the prettiest thing.
The hardest part is saying no to everyone who just wants the pretty thing.
1
2
30
Thanks boss, looking forward to my upcoming pay raise
1
0
29
@RatOrthodox
That's step 1, step 2 is to fund the developer to constantly output new DLC. Just to keep everyone entertained!
0
0
28
@daveloach2
@ByYourLogic
Rest In Peace, he taught me it was okay to work for a company that fixed bread prices
0
0
24
@exhaze
I like how you judge me by 140 characters but don't trust my judgement of somebody after 45 minutes of pair programming
1
1
26
@HipCityReg
@pmarca
Every TMT MD is forwarding that article to their analysts and associates, saying “look, A16Z is becoming like us. Please don’t leave”
0
0
24
It should be illegal for Slack to use their notification sound in tv ads, I am here to relax
1
3
25
these dudes are so cool. "10M context Gemma" - but no eval results and nobody on the Github or HF has managed to run the code properly. key parts of implementation are "left to the reader."
55k people downloaded this model and not a single positive thing to say?
Introducing Gemma with a 10M context window
We feature:
• 1250x context length of base Gemma
• Requires less than 32GB of memory
• Infini-attention + activation compression
Check us out on:
• 🤗:
• GitHub:
• Technical
44
149
1K
3
1
24
You either die a data scientist, or live long enough to become a product manager
2
0
24
The definitive post on Twitter’s potential subscription product
2
0
23
"What's a nice girl like you still doing on the market?" - guy browsing streeteasy, trying to figure out what's wrong with an apartment "listed 3 days ago"
0
1
23
For more details, see our technical report:
Apache-2 weights:
Demo:
1
3
22
This is starting to get into weird territory: added a markdown renderer to the cells and more LLM backends
I intend to open source this (eventually), but come try it out:
Inference via Llama 3.1-8B or 4o-mini is included for free
real spreadsheet, streaming autofill, custom in-line commands, time to do some real work :)
I made this so I could generate and clean lots of high quality synthetic data quickly, and sheets are the nicest UX for that.
1
0
12
7
1
22
HuBERT operates at 50hz (tokens/sec); other labs have reported high quality audio reconstruction is very difficult with fewer than 50hz. This is an unusable token rate: just 1 minute of audio equates to 3k tokens, requiring tons of memory and slower inference.
OpenAI’s new ‘head
3
1
22
My take on useful AI products:
The useful AI apps today are new interfaces or copilots to old interfaces. These can be great businesses but not great venture bets (too high risk or low cap). The "good" bets are full automation agentic plays, because of the upside, but the models
2
1
26
Was inspired by the
@AnthropicAI
test case generator, so I made my own AI spreadsheet. Given some existing values, can we 'fill in the blanks' in a new row? Yes!
This feels like a super intuitive experience to me - what do you think?
1
0
21
Borat is a good reminder that it’s not funny to punch down at others. This is why I no longer make fun of venture capitalists
0
1
21
hey man glad you enjoyed my post, you should subscribe to my newsletter so you can copy more paragraphs without attribution!
1
2
20
Been a long week. Gonna watch relax and unwind by watching Uncut Gems
0
0
20
If you didn’t study 80 hrs a week from age 13 to 22, you won’t have a good enough GPA, from a good enough school, to have the privilege of working 80 hrs a week for a VC
2
0
20
The new models apparently run on HGX H200. In FP8, batch 1, perfect MBU, you can serve up to a 20B dense model at 200 tok/s. With MoE, maybe this is like 40B active params or 80B actives? (not as familiar with MoE inference math).
the latency of the new gpt-4o model is insane - its running at close to 180 - 200 tokens per second for text output. time to first token is also near instantaneous with thousands of input tokens. openai did an amazing job - congrats to the team.
2
3
18
3
0
19
'pickleball but basketball' (hoop is a foot lower so normal people can dunk) would go very hard this summer
1
0
20
Broke: dunking on "an ml engineer turned VC" who doesn't understand basic math
Woke: pitching them your worst startup ideas because they don't understand basic math
1
1
20
Results were inconclusive, let's try again in 2019
1
0
19
Very surprised that a guy who runs a startup based on the economic desperation of the poor doesn’t understand marginal utility
1
2
19
Fast voice to voice! interruptions remain the challenge (and some zero crossing artifacts)
Join us live tomorrow at 2:30pm CET for some exciting updates on our research!
14
42
257
3
0
19
Buddhism was invented by Google engineers in 2009, trying to justify their rest and vest schedules
3
2
19
I deleted all my IDEs because I wanted to write my code the old fashioned way: by paying a guy in India
0
0
19
Thank you for moving humanity forward! Unless you’re a hot young girl, in which case, stop distracting the boys
2
1
19
I Am Once Again Asking You to shut the fuck about statistics and epidemiology if you were a political science major
2
0
18
I used to think the quintessential tech experience was working in SF, surrounded by hip startups, now I understand it to be not working, to be at Barry's midday, to write a blog for 20 readers. At least I got one of those down!
0
0
18
@maksym_andr
Unfortunately I believe the authors have been confused by the ChatGPT ui and simply saw the result of Whisper ASR, NOT anything to do with the audio input modality.
I've replicated figure 7 here using text only input. Speaking gets the same result because... it's just
2
0
18
@_ontologic
You just know somebody is cutting this and tossing it with a generic "China rapprochement bad" IL for politics
1
0
18
As a professional machine learning guy, I'd bet on astrology first
Astrology doesn’t work, but machine learning might.
Suppose you are Facebook or LinkedIn. You have a massive database of life histories. So you could probably do a decent forecast of where a 30-year-old with X job in Y city is likely to be in 5 years, using similar profiles.
197
273
2K
1
0
17
Anybody who says this has never met a counter strike pro. Obsession comes at a cost
If you ever meet someone who at one point was globally ranked in a video game
hire them immediately
826
2K
45K
4
0
18
I trained a SOTA LLM, beating Llama, for less than $500! Here's how I did it: 🧵
LLMs are officially in the web3 era
- undergrads prominently flaunting Stanford creds
- non technical guys writing white papers brought on to "promote and market"
- copy pasted code and weights
Only thing missing was exit liquidity
3
0
69
4
0
17
Brown rice isn't healthier than white rice. It just tastes worse so you don't eat as much
0
1
17