Adya Pandey

AI + enterprise tech intelligence at @cbinsights | prev @michiganross @multitudesworld @fullcycle @bravainvests

[Adya Pandey @AdyaPandey_]

@MenloVentures @AssortHealth @_JonWang @_jeff_liu @mmurph @jpsanday Awesome! We just recognized Assort Health in our AI 100 list 🎉🎉 https://t.co/B9IiVkDHHc

[“paula” @paularambles]

we are weeks away from a startup called Mid raising $100M

[Adya Pandey @AdyaPandey_]

@melisatokmak @Netic_AI Congratulations!

[Adya Pandey @AdyaPandey_]

@NaveenGRao Congrats!!

[Maya Sulkin @SulkinMaya]

Billboards leaving SFO: Stripe, Canva, Figma, etc etc etc

Billboards leaving EWR: Were you injured in New Jersey????

[Meg ✦ @Flicky0ps]

this is what instagram was supposed to be. we lost it to the algorithms and reels and ads and celebs.

[Chris Frantz @frantzfries]

my god what I would do for a Silicon Valley reboot right now

[Adya Pandey @AdyaPandey_]

@agarfinks @WillHurd Congrats!

[Adya Pandey @AdyaPandey_]

ICYMI - the 9th annual @CBinsights AI 100 list is live! See the 100 most promising AI startups building across infrastructure, horizontal, and vertical applications below —

Personal favs: @aaruHQ, @iconicgamesio, @crewAIInc, @PatronusAI, @PhysicalAI ⚙️😃

[Adya Pandey @AdyaPandey_]

@jerseejess @Work_Bench Congrats!!

[Celine Halioua @celinehalioua]

feeling self conscious about my love of em-dashes

[Aaron Ng @localghost]

There's so much content for an HBO Silicon Valley 2

[Justine Moore @venturetwins]

Steve Harvey being chased by monsters in the woods is my new favorite genre of AI images

[Chief AI Officer @chiefaioffice]

The LLM Infra/Ops Market is hot 🔥

“LLMOps” is everything that a dev needs to make LLMs work — development, deployment & maintenance

ChatGPT wouldn't be possible without it

Here's a nice market map from @CBinsights

5 issues these tools help solve:

1⃣ LLMs are expensive

They require constant experimentation and re-training on new datasets (OpenAI GPT model was initially trained on data until 2021) to prevent the model from getting stale.

More importantly, it is the inference costs that are steep (Google can lose $30B)!

2⃣ Fine-tuning is hard

Only a few companies are mature enough to continually fine-tune their models and keep their data pipelines healthy.

LLMs for all their goodness can hence become an architectural nightmare in production — you cannot just train once and deploy-forever.

3⃣ LLMs hallucinate

This is a fancy way of saying — LLMs can lie! This is a major concern as it can spread vast amounts of misinformation (even ChatGPT)

4⃣ Scale and latency:

The real challenge is solving the massive scale requirements of modern applications. Imagine having to train and deploy an LLM in a distributed setting with caching, throttling, and authn/authz, etc. and other critical enterprise features.

Further, trying to understand why hallucinations happen is hard because the way LLMs derive their output is a black box.

5⃣ Privacy and security

We have seen multiple instances of security concerns over LLMs (Eg: Samsung leak).

Prompt Injection is also popular & effective to bypass the rudimentary security of LLMs. Enterprise adoption will not gain steam without much stronger security measures

💡 Wanna hear about the latest AI startups?

Join my newsletter to follow every venture deal in AI (link in bio)

[Will Manidis @WillManidis]

sf tech bros will move to new york and invite six cal grads who work at palantir to their east village rooftop and be like “i love this city’s energy”

[Ana Mostarac @anammostarac]

Blaming my attitude on microplastics

[Andrew Yeung @andruyeung]

New York City puts me in a state I like to call “heightened aliveness”

Grateful to be here.

[CB Insights @CBinsights]

Our experts are going live to explain the #AI landscape, and the impact #genAI has had so far in 2023. https://t.co/b344Vp8gda

[CB Insights @CBinsights]

Watch our analysts break down this year's #AI100 and explain how these companies are shaping the future. https://t.co/ZWcFy9ISWH