HN user

danielcampos93

449 karma

Perfection is the enemy of good https://spacemanidol.github.io/

Posts25
Comments90
View on HN
www.wsj.com 3mo ago

The Decadelong Feud Shaping the Future of AI

danielcampos93
3pts0
semianalysis.com 10mo ago

XAI likely first AI company with Gigawatt plus campus

danielcampos93
9pts0
www.bloomberg.com 10mo ago

What If We're Doing AI All Wrong?

danielcampos93
5pts2
www.ft.com 1y ago

AI talent wars lead to superstar salaries for top tech staff

danielcampos93
1pts0
www.wsj.com 1y ago

The Tech Industry Is Huge–and Europe's Share of It Is Small

danielcampos93
2pts0
www.bloomberg.com 1y ago

The Snowflake-Databricks Rivalry, and Why Both Fear Microsoft

danielcampos93
2pts0
www.bloomberg.com 2y ago

Microsoft Bing Chief Exiting Role After Suleyman Named AI Leader

danielcampos93
2pts1
www.bloomberg.com 2y ago

AI Replaced the Metaverse as Zuckerberg's Top Priority

danielcampos93
4pts0
techcrunch.com 2y ago

Rabbit R1 – Language Action Model

danielcampos93
2pts0
www.bloomberg.com 2y ago

Essential AI Comes Out of Stealth with $57M in Funding

danielcampos93
5pts1
medium.com 2y ago

Evaluating Text to SQL Models Without Execution Accuracy

danielcampos93
2pts0
techcrunch.com 2y ago

Snowflake Launches GenAI tools and copilot

danielcampos93
1pts0
reka.ai 2y ago

REKA releases Yasa, multimodal foundational model

danielcampos93
3pts2
aigrant.com 2y ago

AI Grant 2 Batch Announced

danielcampos93
2pts0
www.snowflake.com 3y ago

Neeva acquired by Snowflake

danielcampos93
166pts111
techcrunch.com 3y ago

Neeva launches its generative AI search engine internationally

danielcampos93
3pts0
news.ycombinator.com 4y ago

Apple Vehicle announcements coming in late 2023

danielcampos93
1pts1
twitter.com 4y ago

Meta scales Language Model to 1.1T parameters using Mixture of Experts

danielcampos93
2pts0
twitter.com 4y ago

Facebook Is Now Meta

danielcampos93
3pts1
huggingface.co 4y ago

Hugging Face Announces Infinity:Ultra-Fast Inference in Your Own Infrastructure

danielcampos93
2pts0
www.reddit.com 4y ago

Faster and smaller Hugging Face BERT on CPU via compound sparsification

danielcampos93
2pts1
www.geekwire.com 6y ago

Xnor.ai Acquired by Apple for 200m

danielcampos93
9pts0
science.sciencemag.org 7y ago

Farming Reshaped Our Smiles and Our Speech

danielcampos93
2pts0
www.1843magazine.com 7y ago

DeepMind and Google: the battle to control artificial intelligence

danielcampos93
202pts137
news.ycombinator.com 9y ago

Ask HN: What Keeps you from moving?

danielcampos93
20pts48

I would love to know what the increased token count is across these models for the benchmarks. I find the models continue to get better but as they do their token usage also does. Aka is model doing better or reasoning for longer?

does it do the full 100? In my experience anything around many items that needs to be exhaustive (all states, all fortune 100) tends to miss a few.

It's not free because it's cheap for them to run. It's free because they are burning that late-stage VC dollars. Despite what you might believe if you only follow them on twitter the biggest input to their product, aka a search index, is mostly based on brave/bing/serpAPI and those numbers are pretty tight. Big expectations for ads will determine what the company does.

This works better because it gives a secondary set of conditions for which the decoder (generating text) is conditioning its generation. Assume instead of their demo you are doing Speech2Text for Oncologists. Out of the Box Whisper is terrible because the words are new and rare, especially in YouTube videos. If you just run ASR through it and run NER, it will generate regular words over cancer names. Instead, if you condition generation on topical entities the generation space is constrained and performance will improve. Especially when you can tell the model what all the drug names are because you have a list (https://www.cancerresearchuk.org/about-cancer/treatment/drug...)

Not mentioned in their blog posts but on the model cards on huggingface: "Molmo 72B is based on Qwen2-72B and uses OpenAI CLIP as vision backbone. Molmo-72B achieves the highest academic benchmark score and ranks second on human evaluation, just slightly behind GPT-4o." Others are based on Qwen 7B. What happened to the Olmo chain?

Great paper! I have a question on how much this can be attributed to portions of text which GPT has already seen. Like if the same approach was followed on a new sliver of Wikipedia which GPT had not seen would the results be the same?

Most other search engines train with a target of google or with some form of reward which is bootstrapped on google rankings. It makes Bing results implicitly have the same behavior as Google. DDG and others just use BingAPI so googles incentives pass on through.