AI WARS HEAT UP: OpenAI, Google, Adobe in EPIC BATTLE!

Welcome Humans🤖,

Here is what we have Today:

• 💡 OpenAI's origin story challenged - Who really invented it?

• 🧠 Math Fails and Word Games: Apple Exposes LLM's Achilles' Heel

• ☢️ Google goes nuclear for AI

• 🎬 Adobe's AI creates Hollywood-level videos

• 💼 New Job Opportunities

Apple

🧠 AI's Reasoning Skills Under Fire

Apple tested over 20 Large Language Models (LLMs) and found major flaws in their logical reasoning abilities including top models from OpenAI.

Which models were tested?

• OpenAI's GPT-4 and GPT-4o

• Google's Gemma 2

• Meta's Llama 3

• And many others

How did they test the LLMs?

• Developed a new benchmark called GSM-Symbolic

• It modified existing GSM8K questions by changing variables

• Added irrelevant information or altered names and numbers

• Focused on evaluating mathematical reasoning skills

What were the key findings?

• Accuracy dropped by up to 65% with minor wording changes

• Even small changes, like altering a name, degraded performance by 10%

• Performance variability increased with question complexity

• Apple concluded there is "no formal reasoning" in LLMs

Are LLMs just pattern matching?

• The study suggests LLM behavior is more likely sophisticated pattern matching

• This challenges the notion that these models can truly reason

Why does this matter?

• Raises questions about LLM reliability in complex applications

• Questions the true 'intelligence' of current AI models

• Challenges the perceived capabilities of leading AI models

What's next for AI reasoning?

• Further research needed to understand LLM limitations

• Potential focus on developing more robust reasoning capabilities

• Possible reassessment of AI deployment in certain fields

This revelation could reshape our understanding of AI's current capabilities and limitations, potentially influencing future AI development strategies across the industry.

Together with AI Tool Report

There’s a reason 400,000 professionals read this daily.

Join The AI Report, trusted by 400,000+ professionals at Google, Microsoft, and OpenAI. Get daily insights, tools, and strategies to master practical AI skills that drive results.

Sign up now for free and work smarter, not harder.

Trending Today

• Adobe just announced the addition of new video generation capabilities — cinematic video, 2D and 3D animations, text graphics, b-roll, and screen effects to blend with normal footage — to its Firefly AI model and Premiere Pro at the company’s MAX Conference, alongside a slew of major AI updates across its creative software ecosystem.

• OpenAI is reportedly involved in a trademark dispute with Guy Ravine, who owns the ‘Open AI’ (with a space) trademark and domain open.ai since March 2015, which Sam Altman and Greg Brockman tried to purchase from him and Ravine claims he conceived and pitched the idea for the initiative to major tech leaders before before OpenAI's launch in December 2015.

• University of Geneva, University of Edinburgh Researchers, and Microsoft developed DIAMOND, an AI model that uses a diffusion-based approach, predicting the next frame based on previous frames and actions, can generate a playable simulation of Counter-Strike(CS:GO) at 10 frames per second within a neural network.

• Google partnered with nuclear startup Kairos Power to build seven small modular reactors in the US, aiming to supply 500 megawatts of carbon-free electricity for AI data centers by 2030.

• DeepMind is experimenting with a new benchmark that suggests LLMs are getting more accurate overall — but still struggle when analyzing large amounts of data at once.

Recommended Reading
If WE had to recommend other newsletters

AI Tool Report is an Newsletter, we enjoy reading daily. AI Tool Report delivers top techniques on how AI can transform your business filled with practical tips, real-world examples.

A Smart Bear is for people who love thinking about strategy, startups, product, marketing, decision-making, and founder psychology.

Job Opportunities

• Shield AI - Quality Manager - Apply

• Palantir Technologies - Site Reliability Operations Analyst - Commercial - Apply

• Synthesia - Strategy and Operations Associate- Apply

• Glean - Solutions Architect Manager - Apply

• C3 AI - Director/Senior Director, Strategic Solutions - Apply

AI ART

Cute cat and dog looking like good friends and be set in Australia. Imagine them as two buddies basking in the sunlight on a beach, watching the sunset. --v 6.1 --style raw.

Ideas? Comments? Complaints?

We read your emails, comments and poll replies daily.

Get the most important AI, tech, and science news in a free daily email.
New Here? Subscribe!
Sponsorship Slots Open for October and Reach over 800+ active readers. (Now 40% off) 🤯

Until next time, Stay Informed!