Daily Digest: Apple's AI secrets

PLUS: Grok goes open

March 18, 2024

Want to get in front of 100k AI enthusiasts? Work with us here

Hello folks, here’s what we have today;

PICKS

  1. Apple quietly released MM1, a family of multimodal language models. They published a research paper sharing the architecture, training data and insights from building these. The largest model in MM1 is 30B and beats many 80B open-source LLMs in visual tasks.

  2. But it still seems like Apple needs some help with the whole generative AI thing. It is in talks with Google to team up and use Google Gemini behind the scenes for upcoming AI features in the iPhone.🍿Our Summary (also below)

  3. Elon Musk stuck to his word and has openly released Grok-1 from xAI. Grok-1 is a massive (314B parameters) mixture-of-experts model. The release is for the base model without any fine-tuning. The performance of Grok-1 lags a lot behind other open LLMs but that size can open up interesting possibilities with continued training.

from our sponsor

Ever wanted to see a truly global AI landscape that covers more than just Western nations?

Our friends at AIport present The Global GenAI Landscape 2024.

This is the first GenAI study to include four times more nations than similar editions, with 128

generative models from 107 international companies included.

See the full version here and subscribe to AIport for more AI insights!

TOP TOOLS

  • KPU by Maisa AI - A new agent-like framework that increases the reasoning capabilities of LLMs.

  • Entelligence - Making engineering organisations efficient, beyond just perfect code.

  • Abel - Review legal records 10x faster with AI.

  • Flowsage - Create flowcharts in half the time.

  • Podsynth - Test a mixture of LLMs with voice models. (beta)

  • Similarix - Thin AI layer for image search in S3 buckets.

  • Tailor - News that matters to you, summarized.

  • Hacksnack by Kadoa - Select your interests and get relevant HN posts.

NEWS

Unclassifieds - short, sponsored links

QUICK BITES

It seems like Apple needs some help with the whole generative AI thing. It is in talks with Google to team up and use Google Gemini behind the scenes for upcoming AI features in the iPhone.

What is going on here?

Apple might license Google's Gemini instead of using its own AI models.

What does this mean?

This lines up with what Apple's been promising, Apple has been working on AI features for the iOS 18, claiming it will break new ground in AI this year. But behind the scenes, it seems they're turning to outside help. Bloomberg reports that Apple wants to partner with Google and use its Gemini AI for these promised upgrades. Apple also talked to OpenAI about doing the same.

In the past few weeks, we have seen Apple release neat AI stuff like MLX (framework for running ML models on Apple silicon), Keyframer (an image animation tool) and most recently a family of multimodal LLMs called MM1. But this lookout for a partnership signals that Apple isn’t that ahead in its AI research that it can beat OpenAI or Google. If you can’t beat em, join em, right?

Why should I care?

Apple and Google are no strangers to such partnerships. Google pays billions to be the default search engine on iPhone. It’s time Google got some back. Google’s Gemini is already powering Samsung’s Galaxy AI features launched in January 2024.

Apple’s AI strategy is still nonexistent publicly. This partnership could be a way to scramble up to headlines of AI progress. Or this might be a covert move to offload heavy tasks to Gemini and use Apple’s in-house AI for more on-device features.

Ben’s Bites Insights

We have 2 databases that are updated daily which you can access by sharing Ben’s Bites using the link below;

  • All 10k+ links we’ve covered, easily filterable (1 referral)

  • 6k+ AI company funding rounds from Jan 2022, including investors, amounts, stage etc (3 referrals)

Reply

or to participate.