Thanks for reading AI Wine Marketing News.
Every two weeks, we deliver an expertly curated summary of AI news relevant to wine marketers, business owners, researchers, and professionals exploring the intersection of wine marketing and our industry's rapidly evolving AI toolset.
If you find our newsletter useful, please share it with a colleague. If you have a question, ask!
Bryan St. Amant – Editor
IN THIS EDITION
Apple: The Illusion of Thinking
Salesforce AI Research: Study shows AI agents struggle with CRM and confidentiality
Futurism: ChatGPT "Absolutely Wrecked" at Chess by Atari 2600 Console From 1977
MIT Media Lab: Your Brain on ChatGPT
Princeton: How to spot real value in AI — and avoid the snake oil
Apple: The Illusion of Thinking
If you think AI technology is moving fast, listen to the hype associated with large "reasoning" models (LRMs). You'll see hype moves faster than technological innovation.
Putting the brakes on the idea of "AI reasoning", Apple's Machine Learning Research group published "The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity."
TL/DR: The only context in which LRMs outperformed LLMs in solving problems that require "reasoning" was when the tasks posed by researchers fell into the "Goldilocks zone," when business tasks were neither too complex nor too easy.
By comparing LRMs with their standard LLM counterparts under equivalent inference compute, we identify three performance regimes: (1) low-complexity tasks where standard models surprisingly outperform LRMs, (2) medium-complexity tasks where additional thinking in LRMs demonstrates advantage, and (3) high-complexity tasks where both models experience complete collapse.
For science-fiction fans enamored with the idea of "thinking" computers, Apple's claims are a buzz-kill.
But for wine marketers relying on ChatGPT or other LLM assistants to perform basic business tasks, these results should be reassuring: 1) you don't typically need a "smarter" AI assistant; and 2) you shouldn't rely on our current crop of AI tools to perform high-complexity tasks.
Read the original paper here: https://machinelearning.apple.com/research/illusion-of-thinking to learn more.
Salesforce AI Research: Study shows AI Agents Struggle with CRM and Confidentiality
Despite the hype surrounding Salesforce's recent launch of "AI Sales Agents," I was surprised to see that they published work highlighting the enormous failure rate associated with using their AI agents to perform even basic automated tasks.
But, good for them! Anyone involved in applying AI theory to automate business workflows should be as transparent as possible when reporting both their successes and failures. That's how we move forward.
I was particularly struck by this result:
“Our extensive experiments reveal that even leading LLM agents achieve only around a 58% success rate in single-turn scenarios, with performance significantly degrading to approximately 35% in multi-turn settings, highlighting challenges in multi-turn reasoning and information acquisition.”
In other words, this finding shows Salesforce's premium-priced AI Agents perform about as well as a new, barely competent intern. Not great news for Salesforce.
However, for wine marketers, these results confirm that our relatively simple use of LLMs to write copy and produce illustrations may be a prudent approach as we learn how to train our new generation of virtual assistants.
If you would like to read the report yourself, it is available here: https://arxiv.org/pdf/2505.18878.
Futurism: ChatGPT "Absolutely Wrecked" at Chess by Atari 2600 Console From 1977
If you need more proof that modern AI isn't the "cure-all" hyped by billion-dollar companies, you'll enjoy this article published by Futurism, "ChatGPT Absolutely Wrecked at Chess by Atari 2600 Console From 1977."
For readers who believe AI is a new invention, it turns out that we've been using artificial intelligence in various forms for more than half a century. As impressive as ChatGPT is, the ChatGPT vs. Atari smackdown reported in this article should remind us all that modern AI hasn't yet cornered the market for "machine intelligence."
For example, wine marketers using ChatGPT to proofread content might find that the algorithms built into tools like Grammarly still have a unique advantage, as they've been trained for many years. Compared to proven tools like this, ChatGPT remains a promising novice that still suffers from too many hallucinations.
Of course, if you want to impress your friends with your chess-playing skills, ask ChatGPT if it's interested in a game of chess.
To learn more, click here: https://futurism.com/atari-beats-chatgpt-chess
Back to the Future?
MIT Media Lab: Your Brain on ChatGPT
Capping off the last two weeks of rough AI press coverage was a study performed by MIT's Media Lab that was routinely publicized with headlines like this:
"MIT study finds that ChatGPT is making people dumber"
The brainwaves of the "ChatGPT" group told a compelling story:
ChatGPT users exhibited significantly less neural connectivity, particularly in alpha and theta waves—signals associated with memory, creativity, and deep thinking.
83% of ChatGPT users couldn't recall what they wrote minutes earlier, while only 10% of essay authors had similar recall issues.
Writing lost its spark. Teachers called the AI-assisted essays "soulless" and "bland."
For wine marketers using AI to write anything, this research highlights the value of consulting our AI assistants prudently if you want to keep your mind sharp.
If you're using AI to speed your creative process, try coming up with ideas on your own first. Then share your ideas with your LLM along with two simple prompts: What am I missing in my first draft? Do you have any other suggestions for reaching our objectives?
I recommend using AI to augment your intelligence, rather than replace it. To learn more, click here: https://www.media.mit.edu/publications/your-brain-on-chatgpt/
Princeton: AI Snake Oil - What Artificial Intelligence Can Do, What It Can't, and How to Tell the Difference
To wrap up this sobering edition of AI Wine Marketing News, I'll point to a new book, "AI Snake Oil: What Artificial Intelligence Can Do, What It Can't, and How to Tell the Difference," written by Princeton Professor Arvind Narayanan.
I agree that AI is overhyped and already causes harm in some applications.
However, as we learn to separate wheat from the chaff, the most productive uses of AI will become routine enough to fade into the background, like any other mainstream technology.
What’s the impact on wine marketers?
During this time of transition, as AI technology continues to evolve, wine marketers open to exploring new ways of creating value have a rare opportunity to advance their careers and businesses by successfully navigating what it means to become an AI-savvy wine marketer.
For a pragmatic and engaging discussion on how to avoid AI snake oil while making this transition, watch this webinar hosted by MIT earlier this year: https://web.mit.edu/webcast/shapingwork/s25/1/
Happy Selling!
ABOUT THE AUTHOR
Bryan St. Amant, founder and CEO of VinterActive, holds an MS in Business Management from M.I.T. and a BS in Mechanical Engineering from U.C. Berkeley. He’s built AI-powered systems for governments, tech giants, and wine marketers. His award-winning work has been featured in books, magazines, and seminars, including CFO Magazine, Inc., CNN Money, eMarketing Magazine, Integrated Direct Marketing, Direct Marketing Association, Grapevine Magazine, Wine Marketing Report, Wine Business, and the Wine Industry Network.