A I C o n t e n t C r e a t o r
Joined December 2024
- Tweets21.6K
- Following62
- Followers8K
- Likes12.2K
Jackson retweeted
she left @harvard surgery to build AI. she was told silicon valley would never accept her, that she would die under a bridge.
so she cold-messaged @KiteVC on linkedin because he backed canva early. he replied the next day and sent her to @SanghaCapital. one 90-minute call later, they were her first investors.
@larsras, co-founder of google maps, came in next. @amandaterry of @ACTAIVentures backed her at pre-seed. then @M13Company and @latifperacha led a $5M seed, with xAI co-founder @ChrSzegedy.
@heyrobynai just launched and ran out of invites on day one.
now her dad thinks they'll IPO in three years.
@iamjennyshao on nontechnical, episode 5.
00:00 she left her harvard surgical residency for a startup
00:46 "leave what? leave surgery? leave harvard?"
01:14 "you're gonna die under a bridge"
03:18 how do you walk away from yourself
05:03 now her dad thinks they'll IPO in three years
06:10 why melanie perkins and canva
07:02 the cold linkedin message to bill tai
07:49 one 90-minute call, then the round
08:32 why people are scared of AI
13:13 "decide who owns the complexity"
16:19 everyone gets their own jarvis
17:37 what if AI takes every job?
19:06 robyn, as in robin to your batman
19:52 "help my dog live longer"
28:45 "as a founder, no one's dying"
30:37 530 builds in one year
32:37 three kinds of surgeons
36:52 why AI forgets everything about you
37:30 three years with nobel laureate eric kandel
40:45 "it's like 50 first dates"
44:11 can AI just order my pizza?
45:19 would you give AI your credit card?
47:59 "jenny, I just need more tokens"
49:50 brian chesky lived with his users. she texts hers.
What would convince us that preference testing needs updating? Ties alone aren't enough. Changes in sensitivity, reproducibility, and relevance would be worth examining.
Drop 2/15
Are we actually measuring the things we ultimately want our voice AI models to achieve?
The best AI voices are becoming hard to tell apart.
From each other. Not from humans.
🚀 Introducing Turing TTS: an evaluation that measures how often human listeners mistake AI-generated speech for real human speech.
We tested leading TTS models against real human speech. Even the best model fooled listeners only ~10% of the time.
With Turing TTS, we hope to contribute to a broader discussion within the research community about how we measure progress in voice AI.
If one of our goals is to develop speech that is indistinguishable from human speech, how well do our current evaluation methods capture progress toward that goal?
We don’t claim that Turing TTS is the right way forward, or that human indistinguishability should be the goal of every TTS system.
Rather, we hope these findings encourage the community to collectively examine whether blind pairwise human evaluations of naturalness are sufficient, and what else we might need to measure.
Results are now live on @voicearena_ai.
Jackson retweeted
We’ve raised $1.3M from YC & Afore Capital so you never have to type or speak again.
Typing is slow. Talking into your computer is awkward. Subvocal fixes both
Our glasses give you the speed of speech without you making a sound or touching a keyboard.
It tracks your internal speech muscles with non-contact sensing and decodes silent, almost invisible mouthing into text at more than 200 words/min.
Pre-orders are now live at:
subvocal.co/preorder?utm_sou…
Jackson retweeted
Introducing Cartesia Managed Agents: our end-to-end voice agent platform for enterprise.
#1 on @ArtificialAnlys Speech Agent Arena and @ServiceNow EVA-Bench, built by co-designing our voice models and harness as one system.
Check it out → cartesia.ai/blog/managed-age…
Jackson retweeted
All that said, I'm generally excited about this new paradigm.
It gives us a whole new axis to scale AI models - towards personalized responses.
Which is more impactful to large scale adoption than just intelligence.
If AI can understand you, it can make things more beautiful for you. We are hoping to help make a more beautiful world for everyone.
Jackson retweeted
BCI is moving away from clinical applications to consumer uses, and applications like these will become more abundant w time.
Robust and dependable regulation on user safety and standards of use on neural data can do great work in setting such standards for safety that allow users to trust such systems.
If brain signals can be used to understand private thoughts and personal preferences, users will want the service providers to be more trustable, and sufficient regulation to enforce the same.
Jackson retweeted
We wanted to talk about this publicly and share how we're thinking about responsibility that comes with such capabilities.
We are focused on how to have higher standards on user consent and are investing in edge inference, and training on encrypted neural data.
We have some interesting work here that I can’t wait to talk more about.
Jackson retweeted
We wanted to be careful about this release given the societal impact of AI that can change faces.
It can cause negative impacts like amplifying body dysmorphia, and narrowing beauty standards, can allow a bad actor to actively influence audience feedback.
Jackson retweeted
Here’s how that works on altering faces based on personal preferences learnt using neural feedback.
Jackson retweeted
Unlike DPO which is generally used to tune text-to-image models, our paradigm is on-policy and doesnt require a preference dataset.
We show that even a simple embedding + k-NN based reward model works well so no explicit reward model has to be trained.
The models move personal preference more strongly than population level preference showing individual preference is captured from EEG.
We also see seed stability across runs where the model converges to similar levels across seeds.
Randomizing the rewards shows that the rewards don’t move, indicating that the reward signal does come from EEG and is not an artifact of training.
Jackson retweeted
In most aesthetic areas, personal preferences deviate quite a lot from population level preferences that internet scale AI models converge to.
This leaves people with poor satisfaction when trying to use these models for tasks that require taste.
So far, there was no way to solve this since getting reliable preference measures from every user is not possible at scale.
Jackson retweeted
How do you get an AI model to make something beautiful for you?
Today non-verifiable capabilities are tuned with either rubric rewards or reward models trained from human feedback, both of them are lossy.
We have finally been able to use arousal signals from the brain to create implicit rewards for personal preferences.
This means personal feedback from each user can be gathered passively with a BCI wearable without explicit user input.
More on the work here
sabi.com/blog/measuring-tast…
Jackson retweeted
Introducing RL with Neural Feedback (RLNF). We have found reliable ways to train AI models to respond to human preferences purely from brain signals.
The example below is a picture of a face that self-transforms into a more attractive version for the subject as they are looking at it.
Real everyday audio with background noise and cheap phones. Training on that seems more realistic than studio recordings.
Jackson retweeted
Oxide just raised $445M at a $6B valuation.
@bcantrill and @sdtuck started @oxidecomputer to build the computers the hyperscalers keep for themselves. @EclipseVentures led the seed when nobody else would. now they're profitable, and he says they didn't even need to raise.
i also asked him if AI is a bubble, why america hates data centers, and why every oxide employee makes the same salary.
nontechnical, episode 4.
00:00 $445M at $6B, and they didn't need it
03:06 one wire transfer bigger than their whole seed
07:14 "they don't have the guts. they're too chicken."
09:09 100% banked at SVB
13:04 a 92-year-old kicking their asses in board meetings
14:48 "would you like me to name names?"
16:58 a16z loved it, then sent the first rejection
20:19 leaving a first round partner meeting for the A's
22:52 "Marc Andreessen was offended"
23:05 NEA: "we will only invest in software"
25:55 "legends are dead"
28:32 "when a VC says 'I've got a technical question'"
42:39 the firm that walked away from a signed term sheet
46:48 "if I could get sequoia, do you think you'd be here?"
49:41 "they were cowards"
52:36 the nuke email he wrote and never sent
58:33 four years later, they asked for a reference
1:02:01 "politics and business don't mix"
1:10:51 is AI a bubble?
1:14:56 how he ruined a dinner party in 1997
1:20:44 "people hate data centers"
1:22:08 how sun got thrown through the windshield in 2000
1:26:26 why everyone at oxide makes the same salary
1:30:58 the worst hire of all time
1:40:42 when oxide goes public
Jackson retweeted
Text-to-speech gets you close. But close isn’t how you’d say it.
So say it yourself, once. HeyGen Voice STS carries the whole performance over to your avatar, down to every pause, emphasis, and whisper.
Test it here: platform.heygen.com/speech-t…
Jackson retweeted
🏆 NVIDIA Nemotron 3.5 Transcribe debuts at #1 among 14 systems on Voice Arena’s Diarization + ASR leaderboard.
Across the full benchmark, it achieves 32.43% cpWER.
This track tests whether a system can get both the words and the speakers right. We evaluated 139 held-out conversations with 2-8 speakers across in-person and online settings, including overlapping speech.
Nemotron 3.5 Transcribe is first in both settings: 34.08% cpWER in person (far field) and 22.13% on online calls.
It has the lowest word deletion rate on the leaderboard. Deletions are 8.32% of reference words, against a median near 31%.
Congratulations to @NVIDIAAI on the release ! 👏
Jackson retweeted
Your OCR says it’s done—but did it get everything right? 🤔 In high-stakes workflows, even a small error can be costly.
🚀 Meet the MiniCPM-OCR suite: an open-source solution to parse fast and guard the results.
Two new models:
🔹 MiniCPM-OCR 1.0 — SOTA OCR performance and speed at 1B scale. Compact, accurate, and fast.
🔹 MiniCPM-OCR Jev — A 2B decision model that helps assess parsing accuracy, with performance approaching that of Qwen3.5-397B-A17B on JevOCRBench.
⚡ What’s new?
Small model, SOTA results: MiniCPM-OCR 1.0 scores 93.68 on Real5-OmniDocBench and 54.7 on OCRBench v2.
A sharp goalkeeper 🛡️: MiniCPM-OCR Jev is model-agnostic, checking text, formulas, and tables from different OCR parsers. It achieves 93.6 on JevOCRBench, outperforming JEV-35B-A3B (90.1) and approaching Qwen3.5-397B-A17B (93.8).
Up to 128 pages per minute 🤯: With DSpark, MiniCPM-OCR runs at 1.49× the speed of PaddleOCR-VL-1.6 and 3.62× that of MinerU2.5 Pro. The SDK’s concurrency optimizations further boost throughput to as much as 128 pages per minute.
Deploy where you work: Day-0 support for Arm, Houmo AI, Intel, and NVIDIA, plus 10 more chip vendors through FlagOS.
Ready for your stack: vLLM, SGLang, llama.cpp, and Ollama, with quantized models, fine-tuning support, and Agent Skills.
🔗 GitHub: github.com/OpenBMB/MiniCPM-O…
🤗 Hugging Face:
MiniCPM-OCR 1.0：huggingface.co/openbmb/MiniC…
MiniCPM-OCR Jev：huggingface.co/openbmb/MiniC…
👉 Try the Demo：platform.modelbest.ai/consol…
Use the API：platform.modelbest.ai/consol…
More subscription time can mean more chances to test what works for your workflow. Flova 1.5's launch promotion gives you a second year of Pro free.
Seedance 2.5 720p starts at $0.025/sec.
flova.ai/?refCode=3BFU8P57
