• AI Academy
  • Posts
  • 🤖 AI's Reality Beyond The Headlines

🤖 AI's Reality Beyond The Headlines

Plus: Amazon's $4B Investment, and Google's Memory Update

November 28, 2024

Hello AI Enthusiast,

This week, we're doing something different. Instead of our usual main news commenting, we'll dive into Benedict Evans' latest presentation where he explores the strategic trends and realities of AI adoption. We think his analysis offers valuable insights worth sharing with you, especially since many of his points echo themes we've been discussing in our newsletters throughout the year.

You'll still find your weekly dose of AI news in the Bits and Bobs section below.

The Big Picture 🔊

Understanding AI's Place in Tech History

Benedict Evans is a tech industry analyst known for his sharp insights on technology trends. Every year, he creates a presentation that looks at the big shifts in tech and what they mean for business and society. These presentations have become a must-watch in the tech world, as they help make sense of complex industry changes by connecting dots between current events and historical patterns. His latest, "AI eats the world", examines AI's impact through this same lens, comparing it to previous tech revolutions while highlighting what makes this moment unique. You can find the full presentation here.

As we do every week, we discussed it internally, and each of us picked one or two key insights from the presentation that particularly resonated with us. Here's what caught our attention and why we think it matters.

These two slides highlight a strategic shift in AI, from Google's leaked memo admitting "we have no moat" in AI to Meta's aggressive open-source strategy, which aims to turn AI models into commodity infrastructure.

James Varnham
James VarnhamCEO and Rainmaker

Meta is following the classic tech playbook of "commoditize your complement", just as Google made Android open source to drive Google Search usage, Meta is open-sourcing AI models to undermine competitors trying to monetize them.

Whether this strategy connects to Meta's metaverse ambitions isn't clear yet.

This slide shows Accenture's rapidly growing quarterly revenue from generative AI projects, reaching nearly $1B by August 2024, while noting that most of these projects are still in pilot phase.

Helin Yontar
Helin YontarCPO and Polyglot

This graph perfectly captures what we're seeing in the field. During my conversations with +130 company representatives this year, I've noticed a common thread: while everyone is rushing to have an "AI strategy," most organizations are still in the experimental phase.

They know they need to do something with AI, but they're not quite sure what that something should be.

This slide explores how LLMs might change the traditional software model, where startups identify use cases, to one where users themselves discover applications through direct interaction with AI.

Gianluca Mauro
Gianluca MauroFounder and AI Rockstar

Think of ChatGPT like Excel, while a spreadsheet can technically be used as a CRM or database, companies typically choose specialized tools for these needs.

Similarly, while ChatGPT can handle many tasks, many business cases require customized AI solutions with proper prompt engineering rather than relying on general-purpose AI tools.

Speaking of specialized AI solutions, our next GenAI Project Bootcamp starts in January. In just seven weeks, you'll learn how to develop customized AI prototypes that address specific business needs. Through hands-on practice with prompt engineering and no-code implementation, you'll master the skills to design solutions that go beyond generic AI tools.

These slides tell a story about AI adoption and automation: from ChatGPT's varied usage across countries to the historical example of elevator operators becoming obsolete.

Gioele Mottarlini
Gioele MottarliniCOO and Image Addict

Despite the hype, AI adoption remains low, with Americans more willing to experiment while Europeans show more caution, probably due to privacy.

But just like automatic elevators, once AI finds its truly useful applications, it'll become invisible, we won't even call it AI anymore.

These slides contrast the traditional "launch now, monetize later" software model with AI's costly reality, where each new model version, like Llama 4, requires exponentially more computing power.

Andrea Mattiello
Andrea MattielloCM and Outdoor Fanatic

A major shift in tech business models is happening. While Facebook could build first and monetize later, AI's massive training costs force companies to have a clear revenue strategy from the start.

The old approach isn't viable when each model costs 10 times more than its predecessor.

This slide examines whether AI-powered search could replace traditional search engines, showing a ChatGPT conversation that raises critical questions about accuracy, navigation, and the preprocessing needed for reliable search results.

Gianluca Belloni
Gianluca BelloniCMO and Marketing Nomad

While getting a single, clear answer instead of scrolling through search results is appealing, AI search could have concerning consequences.

By favoring larger sources and summarizing information, it risks making smaller businesses and websites invisible in ways that traditional search engines don't.

We tried something different today by focusing on one expert's analysis. Tell us what you think!

Login or Subscribe to participate in polls.

Bits and Bobs 🗞️

  • Anthropic introduced The Model Context Protocol (MCP), an open standard designed to connect AI systems with various data sources seamlessly, replacing fragmented integrations with a unified protocol.

  • Amazon is investing $4 billion in Anthropic, naming AWS its primary training partner and offering exclusive AI benefits to AWS customers.

  • Google's AI, Gemini, has introduced a "remember" feature, allowing users to store personal details for more personalized and consistent interactions.

  • OpenAI has released a new update to its GPT-4o model, enhancing its creative writing abilities for more natural and engaging interactions.

  • Anthropic's latest update for Claude allows paid users to connect their Google Drive, enabling direct attachment of Google Docs and Sheets in conversations.

  • OpenAI is expanding ChatGPT’s Advanced Voice Mode to the web for paying subscribers, allowing users to have voice conversations directly in their browsers.

  • NVIDIA has introduced Fugatto, a generative AI model that uses text and audio inputs to generate and transform sounds, including music, voices, and soundscapes.

On the Podcast 🎧

Our "Behind the Scenes of EU Regulations and AI Policies” series continues! This week, we released an interview with Alberto Pena Fernandez, head of ECAT, exploring how to keep humans in control of AI's future while ensuring European values guide its development.

From Our Channels 🤳

Looks like Benedict Evans isn't the only one with a knack for AI predictions! Check out this TikTok where Gianluca shows how his earlier thoughts on AGI are matching today's reality.

@gianluca.mauro

#stitch with @Gianluca Mauro I love being right ❤️ #Ai #learnontiktok #artificialintelligence #business #machinelearning #product #ux #ent... See more

LOLgorithms 😂

This is what happens when your AI facial recognition works too well.

@9gag

Ghost said "hold up lemme try" (📹 @notnathanqnqn) - #ghost #failed #facialdetection #supernatural #9gag

That's a wrap on our newsletter! Here’s a quick recap before you go:

Catch you next week! 👋