A Daily chronicle of AI Innovations June 07 2024:šŸŽ„Kuaishou launches new Sora rival šŸŽ§ OpenAI reverse engineers the workings of AI models šŸ¤–New Chinese video generation model beats OpenAIā€™s SorašŸ“ŠNvidia is now the second-most valuable company, overtaking Apple šŸ”OpenAI probes GPTā€™s inner workings and AI Jobs

Etienne Dieuned Noumen
7 min readJun 7, 2024

--

A Daily chronicle of AI Innovations June 07th 2024:

šŸŽ§ OpenAI reverse engineers the workings of AI models

šŸ¤– New Chinese video generation model beats OpenAIā€™s Sora

šŸ“Š Nvidia is now the second-most valuable company, overtaking Apple

šŸš€Adobeā€™s launches AEP AI Assistant to help brands master customer

šŸ†The most comprehensive benchmarking & leaderboard for image models is here!

šŸ§¬AI used to predict potential new antibiotics in groundbreaking study

šŸ”’Meta gets EU complaints from 11 countries over use of personal data to train AI

šŸ”The Federal Trade Commission is investigating Microsoft-Inflection AI deal

šŸŽ„ Kuaishou launches new Sora rival

šŸ” AI RESEARCH: OpenAI probes GPTā€™s inner workings

Enjoying these daily updates, listen to our podcast at: https://podcasts.apple.com/ca/podcast/ai-unraveled-latest-ai-news-trends-gpt-gemini-generative/id1684415169

šŸŽ§ OpenAI reverse engineers the workings of AI models

In new research, OpenAI has shared improved methods for finding a large number of ā€œfeaturesā€ ā€” patterns of activity in AI models that are human interpretable. They developed new state-of-the-art methodologies that allow scaling sparse autoencoders to tens of millions of features on frontier AI models.

It demonstrated smooth and predictable scaling, with better returns to scale than prior techniques. And they could find 16 million features in GPT-4. The research also introduces several new metrics for evaluating feature quality.

OpenAI has shared the paper, code, and feature visualizations to foster further exploration.

Why does it matter?

It could bring us closer to understanding the inner workings of neural networks. This interpretability will eventually increase model trustworthiness and steerability.

While OpenAI is being criticized for taking unnecessary risks with AI technology, this shows it is also keen on tackling the risk by making models more explainable.

Source: https://openai.com/index/extracting-concepts-from-gpt-4

šŸ¤– New Chinese video generation model beats OpenAIā€™s Sora

Kuaishou, a Chinese tech company, has introduced Kling, an AI model for video generation. It can make videos up to two minutes long at 1080p resolution and 30 frames per second, vs. Soraā€™s one-minute videos.

Kuaishou claims Kling correctly simulates the physical properties of the real world, including complex motion sequences. Using a diffusion transformer, it can also combine concepts and create fictional scenes, such as a cat driving a car through a busy city.

https://youtu.be/cpnKw9djT38

šŸ“Š Nvidia is now the second-most valuable company, overtaking Apple

Kuaishou, a Chinese tech company, has introduced Kling, an AI model for video generation. It can make videos up to two minutes long at 1080p resolution and 30 frames per second, vs. Soraā€™s one-minute videos.

Kuaishou claims Kling correctly simulates the physical properties of the real world, including complex motion sequences. Using a diffusion transformer, it can also combine concepts and create fictional scenes, such as a cat driving a car through a busy city.

https://youtu.be/cpnKw9djT38

Nvidia rallied to record highs on Wednesday, with itā€™s stock market valuation hitting $3 trillion and overtaking Apple to become the worldā€™s second most valuable company. This comes after Nvidia made a series of major announcements in the past week.

However, Nvidiaā€™s stock has surged 147% so far in 2024, with demand for its top-of-the-line processors far outstripping supply as Big Tech races to build out their AI computing capabilities and dominate the emerging technology.

Microsoft remains the worldā€™s most valuable company, with a market value of approximately $3.15 trillion.

Why does it matter?

Nvidia is now seen as a key company to watch to see how fast AI-powered tech is spreading across the business world, a shift that Nvidiaā€™s founder, Jensen Huang, has declared as the dawn of the ā€œnext industrial revolutionā€.

Source: https://www.bbc.com/news/articles/c6ppqd3j621o

šŸŽ„ Kuaishou launches new Sora rival

Chinese tech firm Kuaishou just introduced KLING, a new text-to-video AI model capable of generating high-quality videos up to 2 minutes long with outputs that appear to rival OpenAIā€™s still-unreleased Sora.
The details:
KLING can produce videos at 1080p resolution with a maximum length of 2 minutes, surpassing the 1-minute Sora videos demoed by OpenAI.
KLINGā€™s demos include realistic outputs like a man eating noodles and scenic shots, as well as surreal clips like animals in clothes.
The model uses a 3D space-time attention system to simulate complex motion and physical interactions that better mimic the real world.
The model is currently available to Chinese-based users as a public demo on the KWAI iOS app.
Why it matters: These generations are even more mind-blowing when you consider that Will Smithā€™s spaghetti-eating abomination was barely a year ago. With users still anxiously waiting for the public release of Sora, other competitors are stepping in ā€” and the AI video landscape looks like itā€™s about to heat up in a major way.

Source: https://x.com/rowancheung/status/1798738564735554047

ByteDanceā€™s AI chip loophole

TikTok parent company ByteDance is renting advanced Nvidia AI chips and using them on U.S. soil, exploiting a loophole to sidestep restrictions on Chinaā€™s AI chip exports.
The details:
Due to national security concerns, the U.S. government prohibits Nvidia from selling AI chips like the A100 and H100 directly to Chinese companies.
The restrictions donā€™t prevent Chinese firms from renting chips for use within the U.S. ā€” ByteDance is allegedly leasing servers with chips from Oracle.
ByteDance reportedly had access to over 1,500 H100 chips and several thousand A100s last month through the Oracle deal.
Other Chinese giants like Alibaba and Tencent are also reportedly exploring similar options, either renting from U.S. providers or setting up US data centers.
Why it matters: The AI race between the U.S. and China is only escalating ā€” and it appears major players are going to get AI chips by any means necessary. While the U.S. tries to stall its rivalā€™s progress with restrictions, it feels like a game of whack-a-mole that wonā€™t stop China from reaching its AI goals.

Source: https://www.theinformation.com/articles/chinas-nvidia-loophole-how-bytedance-got-the-best-ai-chips-despite-u-s-restrictions

šŸ” AI RESEARCH: OpenAI probes GPTā€™s inner workings

OpenAI just released a new paper detailing a method for reverse engineering concepts learned by AI models and better understanding ChatGPTā€™s inner workings.
The details:
The paper was authored by members of the recently disbanded superalignment team, including Ilya Sutskever and Jan Leike.
ā€˜Scaling and Evaluating Sparse Autoencodersā€™ outlines a technique to ID patterns representing specific concepts inside GPT-4.
By using an additional model to probe the larger model, researchers found a way to extract millions of activity patterns for further exploration.
OpenAI released open-source code and a visualization tool, allowing others to explore how different words and phrases activate concepts within models.
Why it matters: Much like Anthropicā€™s recent ā€œGolden Gate Claudeā€ and corresponding research, AI firms are still working to understand whatā€™s truly going on underneath the hood. Cracking AIā€™s black box would be a big step towards better safety, tuning, and controllability of rapidly advancing models.

Source: https://openai.com/index/extracting-concepts-from-gpt-4

New AI Job Opportunities on June 07th 2024:

Harvey ā€” Data Analyst: https://jobs.therundown.ai/jobs/61941124-data-analyst
Meta ā€” Applied AI Research Scientist ā€” Reinforcement Learning: https://jobs.therundown.ai/jobs/61972838-applied-ai-research-scientist-reinforcement-learning
DeepL ā€” Corporate Communications Lead: https://jobs.therundown.ai/jobs/62073947-corporate-communications-lead
Palantir Technologies ā€” Enablement Partner: https://jobs.therundown.ai/jobs/61968165-enablement-partner

What Else Is Happening in AI on June 07th 2024

šŸš€Adobeā€™s launches AEP AI Assistant to help brands master customer

Adobeā€™s Adobe Experience Platform (AEP) is a real-time customer experience management (CXM) solution designed for enterprises. AEP AI Assistant is a gen AI chatbot hooked up to the AEP and an enterpriseā€™s storehouse of advertising and customer data, brand assets, and content collateral (at their direction).

Source: https://venturebeat.com/ai/adobes-new-aep-ai-assistant-is-here-to-help-brands-master-customer-data-and-outreach/

šŸ†The most comprehensive benchmarking & leaderboard for image models is here!

AI image models have achieved near-photographic quality. But how do they compare? Are the open-source alternatives on par with their proprietary counterparts? The Artificial Analysis Text to Image Leaderboard aims to answer these questions with human preference-based rankings.

Source: https://huggingface.co/blog/leaderboard-artificial-analysis2

šŸ§¬AI used to predict potential new antibiotics in groundbreaking study

Scientists used an algorithm to mine the entirety of the microbial diversity that we have on earth ā€” or a huge representation of that ā€” and find almost 1million new molecules encoded or hidden within all that microbial dark matter.

Source: https://www.theguardian.com/society/article/2024/jun/05/ai-antibiotic-resistance

šŸ”’Meta gets EU complaints from 11 countries over use of personal data to train AI

Meta rejected the criticism and referred to a May 22 blog in which it said it uses publicly available online and licensed information to train AI as well as information that people have shared publicly on its products and services.

Source: https://www.reuters.com/technology/meta-gets-11-eu-complaints-over-use-personal-data-train-ai-models-2024-06-06

šŸ”The Federal Trade Commission is investigating Microsoft-Inflection AI deal

The FTC has sent subpoenas to tech giant and startup, asking whether their partnership evaded the required government antitrust review of the transaction. Microsoft picked off Inflectionā€™s specialized workforce of AI researchers but didnā€™t purchase the company outright.

Source: https://www.wsj.com/tech/ai/ftc-opens-antitrust-probe-of-microsoft-ai-deal-29b5169a

Enjoying these daily updates, listen to our podcast at: https://podcasts.apple.com/ca/podcast/ai-unraveled-latest-ai-news-trends-gpt-gemini-generative/id1684415169

--

--

Etienne Dieuned Noumen
Etienne Dieuned Noumen

Written by Etienne Dieuned Noumen

šŸ§Ŗ Senior Software Engineer | Tech Lead šŸš€ AI/ML Enthusiast šŸŒ Canadian with African roots | Proud father of 4 āš½ Lifelong soccer player and coach

No responses yet