FastForward #79: The language we use to describe AI matters
Hi everyone. I’d like to thank Akamai once again for supporting FastForward for the month of September. If you're interested in advertising, please drop me a line at ads@fastforward.blog. If you like what you see, please share this week’s edition with a friend, and encourage them to subscribe. It really helps.💌 Sign up here.
ForwardThinking 🤔
The language we use to describe AI matters
Here we go again. Another AI researcher emerged this week warning us of humanity's impending doom due to the power of AI. This one was a former Anthropic employee who quit his job because he is convinced these models are going to kill us all. I thought we were past the doom-speak, but apparently not.
Alongside these types of pronouncements, which we have been hearing on a regular basis since ChatGPT emerged in 2022, there is a charged language developing around describing this generation's AI capabilities. It seems deliberately designed to be scary and threatening to make the AI software sound more capable than it really is, and attribute actual agency, intelligence and even menace to the AI.
Consider that when describing recent agentic AI-led hacks of Hugging Face and a German wiki, the nomenclature used by OpenAI was "swarms of agents," wording which was picked up and used frequently in the press. Meanwhile, some went so far as to describe agents as actually creating "civilizations of agents." Words like this raise images of the zombie apocalypse movies and dystopian science fiction instead of what podcaster and Georgetown professor Cal Newport described as merely a way to deal with complex prompts.
"So what's going on with an agent swarm? It's not about some sort of exotic type of intelligence. It's prompt management,” Newport said in a recent episode of his Deep Questions podcast. "So this is a swarm in the same sense that having multiple programs open on your Mac is a swarm of programs," he added. (I encourage you to listen to the whole episode for his complete explanation.)
Even relatively benign language like hallucinations, thinking and reasoning assigns human-like capabilities to software that in my experience, while able to do some pretty amazing things, isn't close to the kind of human-like intelligence that we keep hearing attributed to current AI.
That is not to say that recent agent hacking incidents, however you choose to describe them, didn't happen. There's plenty of evidence they did, even if most of it comes from the labs themselves. The issue is the language we use to describe them, and the motivation behind the words.
Real world versus alarmists and true believers
There is definitely a gap between what I call the reality of everyday use and AI true believers, who are ready to accept every researcher's alarmist call that this software is so good, it's capable of destroying civilization. In fact, the more time I spend working with these tools, the greater the disconnect I find between the language the industry is using to describe them and what they actually do.
I spend time with AI every day. I use Claude, ChatGPT, Perplexity and Writer (which gave me a free subscription). These tools have become a regular part of my research, editing and review workflow, and even business document creation.
I've been able to generate a number of documents for the business side of FastForward including proposals, invoices, contracts and marketing documents; all from a description, which would have taken hours before. But it doesn't just magically happen on the first try. It takes a lot of work, sometimes incredibly frustrating work, to get it done. These models backslide. They forget. They make dumb formatting mistakes and struggle to fix them. They are not all-knowing or anywhere close to human-like intelligence.
More recently, I've tried to teach a Claude agent to become my news gatherer for this newsletter. Each day I check the Techmeme headlines and choose some of them for news of the week, commentary I'm considering, what I'm reading and what I'm watching. Each morning I review the Techmeme email, make my choices and quiz the agent on what it chose. While it's improving incrementally, after a couple of weeks of back and forth and refinement upon refinement, it's still not really close. This isn't a simple task, but the agent lacks human flexibility. It looks for fixed, rigid rules where none exist and human judgment comes into play. Wouldn't a human-like intelligence pick this up much faster without so much input from an actual human?
I keep coming back to the idea that AI has some astonishing capabilities today. Why do we need to make it sound more powerful and actually panic people about it? The labs have their agendas, of course, to make the models sound bigger, badder and scarier. Industry insiders appear to be stuck in a feedback loop, reinforcing one another’s perception about the danger of these models. But for the people outside of this bubble, words matter because they can suggest levels of intelligence and agency that everyday use simply doesn't support.
~Ron
What's new on the blog 📰
Nvidia-Hugging Face deal could raise questions about AI power consolidation
It's not exactly a secret that Nvidia is the richest company in the world, and if the Hugging Face deal passes regulatory muster, it will only add to its growing clout in the AI market. I suggest that perhaps regulators should take a closer look at this one.
Note that I pulled this article from last week's News of the week section in FastForward #78 and published it with light editing.
How Kyndryl’s separation from IBM prepared it for the AI era
I recently interviewed Kyndryl CIO Kim Basile about what she learned as the company separated from IBM a few years ago. And the really interesting thing was that it forced them to modernize their systems and get their data house in order, and without even knowing it at the time, prepare them for the AI age. It's a pretty fascinating story.
As she told me, the separation ended up preparing Kyndryl for what came next:
"It gave us a foundation and many companies are still trying to build that foundation."
How a 150-year-old insurance company is moving into the AI era
I interviewed MetLife CIO Nick Nadgauda to learn how a 150-year-old insurance company is transforming its tech stack for the AI age.
Using COBOL programs as an example, he says he looks at what to transform more from a maintainability perspective than a strict modernization one.
"I don’t necessarily have a problem with COBOL. The real problem in my mind is when you’ve got systems that you can’t support, where you can’t find programmers, you can’t find talent or they’re hard to modify."
Why Cisco’s Jeetu Patel thinks product strategy should drive acquisitions
I had the pleasure of hosting Cisco president and chief product officer Jeetu Patel on my FastForward on PPN podcast recently. I asked Jeetu about his build versus buy strategy, and he told me it was all about the products.
"The reason I don't like having a strategy which says what is your acquisition strategy, is that I don't have an acquisition strategy. I have a product strategy,” Patel told FastForward on PPN.
The entire podcast is embedded in the article.
In the debate over open-weight models, Capital One makes a different case
There is a battle raging in tech over the use of open weight models. On one side, the AI labs, the same ones telling us that their models got loose and hacked several websites, say the open weight models are unsafe.
Other companies argue that they need to be able to customize models, and the proprietary ones from those same AI labs make that much more difficult to do.
I spoke to Milind Naphade, SVP of AI foundations at Capital One, whose company doesn't care about the debate. They just need the flexibility that open weights provide to take advantage of their data.
"To do all that needs open-weight models, we just cannot do that without them. And why is that important? Because that's the only way that we can fulfill our twin mandates of delivering high accuracy with regulatory compliance," he said.
News of the Week 📣
Adobe names Anil Chakravarthy as CEO to replace Shantanu Narayen
Adobe announced this week that Anil Chakravarthy would be the company's next CEO, replacing long-time chief exec Shantanu Narayen, who announced he was stepping down last March after 18 years on the job to become executive chair.
Chakravarthy comes from the company's enterprise side of the house. He was most recently president of Adobe’s customer experience orchestration business and worldwide field operations. Prior to joining Adobe in 2020, he served as CEO at Informatica, so he is not without experience in the corner office.
He said all the right things in a statement announcing his promotion to CEO. "I'm energized by the opportunity to fuel growth and drive Adobe's leadership in the next era of agentic software for creativity, productivity and customer experience." Like his fellow SaaS companies, he's going to have to find a way to move beyond per-seat licensing, shift to agentic workflows and help move creatives into the AI age.
The company is perhaps better known for its creative tools like Photoshop, and another long-time executive David Wadhwani, who ran the creative tools division and was reportedly considered a leading contender for the CEO job, announced on the same day that he would be stepping down later this month and leaving the company after a period of transition.
Some have suggested that there might have been a power struggle between the creative side of the house and the enterprise side, and the board decided the future was with enterprise.
The company reported earnings on Thursday with third quarter revenue of $6.76 billion up 13% YoY. Double-digit growth for a mature SaaS company in the AI era is nothing to sneeze at, but the stock price was down almost 3% at the market open this morning. Perhaps investors are skittish about the leadership transition despite the strong showing.
Chakravarthy takes over starting on December 1st and begins the next chapter for the company.
Qualcomm's custom chip deal with AWS comes with lots of qualifiers
Qualcomm announced a deal this week with AWS, which could be lucrative for both sides, but comes with a ton of qualifiers. The deal for custom chips is for up to 10 years of multi-generational chips and up to $60 billion in value for Qualcomm, but there is no guarantee the full deal will come to fruition.
To sweeten the pot, Qualcomm has issued stock warrants of 25 million shares at $161.26 per share for a total value of $4 billion, per Reuters. The catch here is that these only kick in after certain sales milestones are met, and Amazon can choose to buy them or not. It's worth noting, however, that according to the 8K form filed by Qualcomm with the SEC, Amazon is eligible to buy 3.75 million of those shares immediately if it so chooses, based on initial commitments (although it's not clear just how much that is).
The deal isn't exactly circular like some of the ones we have been hearing about from Nvidia with its customers, but the stock sweetener does highlight the importance of this agreement to Qualcomm, which until recently concentrated mostly on mobile chips. In recent years it has been diversifying into areas like custom data center chips, industrial IoT and automotive.
As AI increases hyperscaler requirements for companies like Amazon, chip makers like Qualcomm can take advantage of the growing demand. "As AI demand accelerates, data center infrastructure will require advances in both computing and connectivity to deliver greater performance with more efficiency," company CEO and president Cristiano Amon said.
The deal, which includes customized silicon for AI inference and optical connectivity technology, comes after agreements with Meta and Microsoft as the company tries to expand its data center-focused business.
Bending Spoons stays on a SaaS bender with Miro acquisition
Bending Spoons likes once highly-valued SaaS companies with revenue. The latest is Miro, once a darling of the pandemic digital software era with a high water valuation of $17.5 billion. Bending Spoons announced it was acquiring them this week for $1.355 billion. With net cash on hand, it brings the equity value of the deal to $1.79 billion.
Miro provided digital whiteboarding at its peak, giving people who longed to be in a conference room together, the ability to brainstorm in an online meeting. While that still holds, the company has modernized and today bills itself as an AI innovation workspace where people can collaborate and maintain context across projects.
The company isn't a distressed property by any means, reporting over 4 million paying users, 750 of which are organizations paying over $100,000 a year. That's not chicken feed and it's just the kind of company that Bending Spoons gravitates towards.
Bending Spoons is a private equity play of sorts, scooping up companies that once had high valuations and buying them for pennies on that high mark. The company did it with Airtable recently and prior to that Evernote, and has bought over 50 companies over the years. It tends to hold onto these companies, cut costs (and employees) and squeeze more profit out of them.
The acquisition is subject to regulatory approval, but is expected to close in the fourth quarter this year.
What I'm reading 📚
Tech explores Argentina's Patagonia for mega data centers
~By Leila Miller, Reuters
The AI Conversation CEOs Aren’t Having With Their Teams
~By Brian Solis, LinkedIN
On AI Coding and Its Discontents
~By Cal Newport, Cal Newport Blog
What I'm watching 📺
The Future of AI-driven Software Delivery | Jyoti Bansal
~FastForward on PPN, Ron Miller
Look who's talking 👄
"We are shipping one big feature every 11 hours right now. I could not imagine that was possible."
~Harness founder and CEO Jyoti Bansal on FastForward on PPN this week.
