Jacob Marks

ML Research @ Liquid AI | Theoretical Physics PhD @ Stanford | ex-Google X, Voxel51

[Jacob Marks @QuantumMarks]

@DataScienceHarp @llm_wizard @GregOnLock too true haha 🤣

[Voxel51 @Voxel51]

Voxel51 announces the close of Series B funding led by @BessemerVP and including @drivecapital, @TopHarvestCap, Tru Arrow Partners, & more! We're thankful for our amazing investors and community, and excited to continue making #visualai a reality.

https://t.co/OpiebmMfI9

#ai

[Jacob Marks @QuantumMarks]

Let's Go!!! @Voxel51 🤝 @MongoDB for the win. Excited to attend MongoDB .local NYC tomorrow!!

[MongoDB @MongoDB]

@anyscalecompute @robertnishihara @freeplay_ai @symbldotai @surbhirathoree By combining @Voxel51's open source computer vision toolkit with Atlas Vector Search, developers can quickly perform vector similarity searches across billions of features with just a few lines of code, both in Python and via pp. Excited to partner with Voxel51!

[Albumentations @albumentations]

If you use @albumentations in your library or product we would be happy to add a page to our docs on how to use @albumentations in synergy with your tool.

Example:
@QuantumMarks wrote

https://t.co/e52DHMFlyq

About integration in @Voxel51

We cross-posted to our docs

[Jason Corso @_JasonCorso_]

🚀 Big news at @Voxel51! We're now hiring AI researchers & scientists to join our new science team.

As FiftyOne has amassed a 2M+ install-base, I’ve had the luxury of interacting with many hundreds of open source and corporate users, learning about significant gaps in building performant AI solutions. Our new science team will work toward rectifying these gaps through foundational innovations, which we will publish and bring to the community through FiftyOne.

All science. All publishable. All open source.

🧠💡 Excited? Learn more and apply now! https://t.co/VkamdhjcfX

[Jacob Marks @QuantumMarks]

Hands-on tutorial on clustering visual datasets using embeddings, with Scikit-learn, FiftyOne, and CLIP!

🔗 https://t.co/ecbljYjPNT

#computervision #ai #ml #DataScience

[Jason Corso @_JasonCorso_]

🔍 New Blog: Observations on MLOps – A Fragmented Mosaic of Mismatched Expectations. A journey into the complexities of #MLOps, exploring the vibrant yet challenging world of AI/ML in production, from a technical academic outsider!

👉 Read more: https://t.co/TMRq5UKCYV

[Jason Corso @_JasonCorso_]

Check me out on FrontLines https://t.co/kGV0czJL85 where we talk about the fun and challenges of operating an AI startup in a /noisy/ time, with an academic background.

[Jacob Marks @QuantumMarks]

OpenAI's CLIP is cool, but have you heard of OpenCLIP, MetaCLIP, Data Filtering Networks, ALIGN, or K-LITE??

Check out today's article on the history of CLIP model training data advances!
🔗 https://t.co/pmLaKTmSaD

[MongoDB @MongoDB]

The vast majority of the world’s data is unstructured. Learn how the integration between @Voxel51 and MongoDB Atlas enables the processing and analysis of visual data across billions of features with just a few lines of code. ⬇️

https://t.co/ewUDTAm4WN

[Jacob Marks @QuantumMarks]

Data augmentation is hard... so test your transformations before training!

Want to know how? Check out my latest post on using @albumentations and FiftyOne to avoid the perils of blind data augmentation!

🔗 https://t.co/e7crYzEQvD

[Jacob Marks @QuantumMarks]

Incredible advice for any researchers trying to stay afloat in the sea of new "advances"!

[Jason Corso @_JasonCorso_]

There were more than 11K research papers in the top machine learning and computer vision conferences in 2023. WOW!

How can you read that many papers? How can you retain what you read? Check out my new article for answers!
https://t.co/6OJU3CmabC

[Jacob Marks @QuantumMarks]

Really provocative piece by UMich computer vision professor @ProfJasonCorso

[Jason Corso @_JasonCorso_]

Annotation is dead.

Human annotation is largely responsible for the current AI boom, but is the need for human annotation gone?

I wrote a long-form article on the future of human annotation in ML https://t.co/SiU35TUj3p

[Jacob Marks @QuantumMarks]

@gabrielaltay Amazing work @gabrielaltay !

[Jacob Marks @QuantumMarks]

NeurIPS 2023 and The State of AI Research

I scraped the data from NeurIPS papers over the past 10 years and did a meta-analysis.

Want to see the trends? Check out the blog post: https://t.co/rYSdiYs7UD

Going to NeurIPS? Say hi! We'll be at booth #427!

#NeurIPS @NeurIPSConf

[Jacob Marks @QuantumMarks]

@patrickmineault Not neuroAI papers in particular, but I made this list of NeurIPS 2023 papers that you won't want to miss: https://t.co/TQ1UnOihgz

[Jacob Marks @QuantumMarks]

@rachelmetz @rachelmetz I'll be there too!

Also, made a collection of some of the papers I think are super cool: https://t.co/TQ1UnOihgz

[Jacob Marks @QuantumMarks]

@NeurIPSConf Abstracts:
- Shortest abstract was 29 words (poster #70813)
- Longest abstract was 373 words (poster #74168)
- 675 abstracts explicitly mention GitHub!

[Jacob Marks @QuantumMarks]

In advance of #NeurIPS2023, I made a repo with
- comprehensive data analysis
- downloadable info for all papers
- and a collection of papers you won't want to miss!

🔗 https://t.co/yXKcSoVD2s

@NeurIPSConf

This thread contains some of the highlights!

[Jacob Marks @QuantumMarks]

@NeurIPSConf Authors:
- # of authors/paper increased from 4.66 to 4.98
- 13,012 unique author names, compared to 9,913 in 2022