📈 How are companies using AI?

➜ Major AI Upgrade for Amazon’s Autonomous Stores. Amazon has developed an advanced AI model for its Just Walk Out checkout technology. Major change: instead of gathering input piece-by-piece from its cameras, shelf sensors, 3-D store model, and product data, the new AI model analyzes all input at the same time to determine exactly what shoppers are picking up, putting down and ultimately walking out with. TL;DR AI makes Amazon’s tech more accurate, faster, and cheaper.

RetailDive

➜ Microsoft Office 365 Copilot Not Worth The Money? A CIO of a pharma company paid an extra to have 500 employees use Office 365 Copilot in the fourth quarter of 2023 and first quarter of 2024. This is Microsoft's much-heralded AI upgrade to its popular suite of productivity software. After six months, the exec canceled the upgrade and compared it to middle-school presentations: "The price is double and we really just do not see the value we're getting out of those tools worth double." Ouch.

BusinessInsider

➜ 10x More Computing Power. Zuckerberg says Meta will need 10x more computing power to train Llama 4 than the company used to train Llama 3: “I’d rather risk building capacity before it is needed rather than too late, given the long lead times for spinning up new inference projects.” Meta’s capital expenditures rose nearly 33% to $8.5 billion in Q2 2024, from $6.4 billion a year earlier – servers, data centers and network infrastructure are a costly business. 

TechCrunch

➜ When AI is Mentioned, People Are Less Likely to Buy. A study from the Journal of Hospitality Marketing & Management found that AI can be a turn-off for consumers: "When AI is mentioned, it tends to lower emotional trust, which in turn decreases purchase intentions," said the study’s lead author. Researchers tested the effect across eight product and service categories getting similar results. It’s not so much about incorporating AI into products – it’s about marketing them with a focus on artificial intelligence. Are we already tired of the hype?

Futurism    

➜ European Artificial Intelligence Act Is Here. The act categorizes AI based on product safety and risk, defining systems such as recommendation engines as ‘minimal risk’, AI generated content as posing ‘transparency risk’, AI systems used for recruitment or credit scoring as ‘high risk’, and systems that enable social scoring as ‘unacceptable risk’. The regulation introduces a set of standards and companies not complying with the rules will be fined. Fines could go up to 7% of the global annual turnover for violations of banned AI applications, up to 3% for violations of other obligations, and up to 1.5% for supplying incorrect information. The rules of the AI Act will start applying on 2 August 2026.

Finextra

💡 What are the nerds up to?

➜ Do Developers Hate Their Jobs? According to a new Stack Overflow survey, software engineers globally embrace AI. As much as 72% of respondents are in favor of leveraging AI tools in their daily work and 81% of them cite increased productivity as the main benefit. What’s more, 70% of developers don’t worry about losing their jobs to AI. However, it turns out that 4 in 5 respondents are unhappy. Main factors include dealing with technical debt, complexity of tech stacks, and reliability of the systems they work with. Experts also point to stress and uncertainty caused by the recent massive layoffs.    

Shiftmag

➜ How to Run Llama on Home Devices? Thanks to Distributed Llama, it’s now possible to run huge LLMs across multiple home devices. All it takes is an AI cluster! Distributed Llama leverages tensor parallism and is optimized for the low amount of data required for synchronization.
Medium

➜ Building Custom AI Agents When You’re Not a Geek. A new tool on the block, Wordware, has just launched their IDE platform for building AI agents which allows domain experts to create custom content generation models, invoice processing, data querying, reporting, and sales enrichment using an LLM backend, without the need to know how to code.
Product Hunt

➜ New Copilot, But Better? Built to make engineers ‘extraordinarily productive’, Cursor looks a lot like GitHub Copilot, with pretty much the same functionality. I’m hearing more and more stories from developers praising Cursor and saying it works much better than the predecessor. 

Cursor

➜ Gemini 1.5 Pro Beats GPT-4. Have you seen Google AI Studio? A 2 million token context window, improved customization, effortless model tuning capabilities, the ability to add video, and code execution (so you can sit back while the tool creates charts from your data) – word is Gemini 1.5 Pro is much better than GPT-4.

Google AI Studio

➜ AI Is Not Overhyped. Now here’s a developer voice with a clear argument. A great longer read on how to make the most of LLMs in the world of engineering by Google DeepMind’s Nicholas Carlini. “I would say I’m at least 50% faster at writing code for both my research projects and my side projects as a result of these models,” claims Nicholas.   

Nicholas Carlini’s blog

Thanks for reading!

I share these stories because I believe it's important for all of us to keep up with AI. To support my mission, share this newsletter on LinkedIn.