Almost Timely News: The Dawn of Autonomous AI (2023-04-23) :: View in Browser

Almost Timely News

👉 Watch my brand new keynote, The Marketing Singularity, all about how generative AI is the end of marketing as we know it »

Content Authenticity Statement

100% of this newsletter was written by me, the human, with no contribution from AI. If there ever is a contribution generated by AI, it will be prominently disclosed.

Watch This Newsletter On YouTube 📺

Almost Timely News: The Dawn of Autonomous AI (2023-04-23)

Click here for the video 📺 version of this newsletter on YouTube »

Click here for an MP3 audio 🎧 only version »

What’s On My Mind: The Dawn of Autonomous AI

This past week was another wild one, this time with the dawn of autonomous AI. Well, that’s not strictly true; we’ve had autonomous AI for a while, but not specifically with large language models and not in ways that are easily accessible. Let’s talk about what this stuff is, what it means, and what you and I should be doing about it.

First, what is autonomous AI? Autonomous AI is AI that does stuff itself. We give it some general directions, and then it goes and does those things. Probably the most well-known type of autonomous AI is the self-driving car. You put in a destination, and the car drives itself, figuring out how to get from where you are to where you want to go. Along the way, it drives, navigates, and communicates how the trip is going, dealing with traffic, pedestrians, etc. We’ve seen plenty of demos about how this sort of technology works, and for the most part, it does work about as well as a human driver – perhaps slightly better. At least the AI driver isn’t staring at its phone while changing lanes at 80 miles an hour on the highway.

We see examples of autonomous AI even within our homes. If you’ve ever gotten one of those smart robot vacuum cleaners, that’s autonomous. Given a programmed schedule and the restrictions you want it to obey (usually programmed by you in an app), it does its thing until either the task is done or it’s devoured an errant power cable on your floor for the third time this week.

Now, in the context of large language models, models like the GPT family from OpenAI, Google PaLM, StabilityAI’s Stable LM, and many others, what does this mean? We’re used to interacting with these models in a singular fashion. Open up an instance of ChatGPT, start typing your prompt, and it does whatever you direct it to do. (assuming it’s in compliance with the terms of service etc.) That’s a single instance of the model within the interface, and for many tasks, that’s good enough.

Suppose you were able to start two instances of ChatGPT. Suppose one instance could hear what the other instance was saying and respond appropriately to it. You’d sign into one instance and tell it to start writing a blog post. You’d sign into the other instance and tell it to correct the blog post for grammatical correctness and factual correctness. Both instances would start almost competing with each other, working with and against each other’s output to create an overall better outcome.

That’s the essence of autonomous AI within the context of large language models. They’re multiple instances of a model working together, sometimes adversarially, sometimes collaboratively, in ways that a single instance of a model can’t do. If you consider a team of content creators within an organization, you might have writers, editors, producers, proofreaders, publishers, etc. Autonomous AI would start up an instance for each of the roles and have them perform their roles. As you would expect in a human organization, some tasks are collaborative and some are adversarial. An editor might review some writing and send the copy back with a bunch of red ink all over the page. A producer might tell the editor they need to change their tone or exclude negative mentions about certain brands or personalities.

So, why would someone want to do this? There are plenty of tasks – complex tasks – that require more than a single prompt or even a series of prompts. They require substantial interaction back and forth to work out key points, to deal with roadblocks, to collaborate and create greater outputs than they could individually. These tasks are the same ones people often work together on to create better outputs than they could individually. I might have an idea I want to write about, but I know that for a significant number of them at work, my ideas get better when I discuss them with Katie or John. Sometimes they behave in a collaborative way, asking “what if” questions and helping me expand on my ideas. Sometimes they behave in an adversarial way, asking “so what” questions and making sure I’ve taken into account multiple points of view and considerations.

That’s what an autonomous AI does. It plays these roles against itself and with itself, working as a team within a computational environment. It’s like an AI office, where the individual office workers are AI instances.

What would this look like as an example? Here’s the setup I devised in AutoGPT, one of the most popular versions of this technology. AutoGPT asks for an overall purpose and five goals to accomplish. Here’s what I told it to do:

  • You are a nonfiction author. You write books about marketing, marketing analytics, marketing attribution, attribution modeling, marketing mix modeling, media mix modeling, media spend, marketing strategy, marketing budgeting. You will write the outline for a book about marketing mix modeling using LASSO regression. You will write in the style and voice of marketing author and expert Christopher S. Penn.
  • The book you will write will be a total of 60,000 words about marketing mix modeling. You will write 20 chapters of 3,000 words per chapter.
  • You will write about why marketing mix modeling is important, what marketing mix modeling is (with examples), and how to implement marketing mix modeling in the R programming language with plenty of examples.
  • You will review your writing to ensure the book is 60,000 words or more, grammatically correct, coherent, and appropriate for business book readers. You will ensure that you have correctly captured the writing style of marketing expert Christopher S. Penn.
  • You will export your work in Markdown format, one Markdown file for each chapter of the book. The book’s author is Christopher Penn. The year of publication is 2023. The publisher is TrustInsights.ai. The book is published in the United States of America.

Once I got the software installed on my laptop, configured, and ready for use, I started up the engine and put in my goals:

AutoGPT

We see above, it’s getting started and working out the structure of what it needs to accomplish. It knows it needs to extract data about what marketing mix modeling is, what my writing style is, and outline the book. About 20 minutes after I issued these commands, it started cranking away:

AutoGPT mid process

These are the first passes through, just getting together the text. It hasn’t started checking over its work to ensure each chapter is the correct length (it’s not), or that it’s coherent and matches my writing style. But you can see just from these examples, from this process, that it’s going to do what I directed it to do in a very efficient way. This is what autonomous AI looks like.

Now, let’s be clear. This isn’t sentience. This isn’t self-awareness. The machine is not alive in any way, shape, or form. It still needed me to declare what it was supposed to be doing. It has no agency of its own without that initial direction, something to kick off the process, so banish any thoughts of Terminators or Skynet. All kinds of folks are talking about this as the start of artificial general intelligence, of truly intelligent artificial life, and it’s not. This is no more alive than a self-driving car. Your cat has more agency than this, more free will. That is not the threat that this technology poses.

What threats does it pose? A few. First, as you can see from the example, this dramatically increases the complexity of tasks that large language models can tackle in a relatively straightforward way. Up until now, large language models struggled to deal with very large forms of text, like novels and books. They don’t generate those well in a singular fashion. This can do so, dealing with far more complex problems and tasks.

Second, this technology exacerbates issues with copyright. At one point, AutoGPT opened up a web browser and started surfing my website to get a sense of my voice and tone. That’s okay – it’s my website, and obviously I give it permission to do so. Suppose I had suggested someone else’s voice instead? That’s problematic, and there’s no ethical checksums, no checks and balances in the technology to say, “hey, maybe don’t do that”. The tool is truly agnostic, truly amoral. It has no concept of right or wrong, which means that any morality needs to come from us.

And that brings us to the third problem. This tool has no morals, good or bad. It only understands the tasks you give it, and it works to achieve those tasks. Morality is in the eye of the beholder. Suppose I wanted the tool to generate some propaganda. Would it do that? Yes, unquestionably. Suppose I wanted the tool to scrape some data from LinkedIn. Would it do that? Yes, yes it would. Suppose I wanted the tool to find a working login to a secured website. Would it do that? Yes, it would. Without going into any details, I asked it to try to break into my personal website, and it went about trying to figure that out. Did it succeed? Not at the time I tried it, which was 5 days ago.

In the last 5 days, the ecosystem around the tool has introduced dozens of plugins that make the tool more capable, like different kinds of web browsing, connections to services and APIs, all sorts of capabilities. It’s a very small stretch of the imagination to envision tasks that autonomous AI could undertake that you might not want it to. People who work in cybersecurity should be very, very concerned and should be watching these kinds of tools like a hawk. They should be red-teaming with these tools today to understand what their capabilities are and are not.

The output right now out of tools like AutoGPT stinks at the moment. It’s coherent but it’s boring, and the process is janky as hell. It’s not ready for prime time…

… just like GPT-2 wasn’t ready for prime time three years ago. And today, GPT-4 and similarly sized models are in production, in the world, and working really, really well at a large number of tasks. Autonomous AI is just getting started, so to dismiss its shoddy output today and assume it will not evolve is just short-sighted.

AutoGPT animation

Whether or not we wanted this technology, it now exists and is available in the world. So what should we do about it?

At a personal or organizational level, we need to be doing rigorous audits of the kinds of work we perform to see what other tasks AI could take on. I’d initially thought that large language models couldn’t easily take on very large content tasks until next year, and here we are. In what ways could you use technology like this for longer-form content like books, keynote addresses, movie scripts, entire publications? Start today doing an audit, then start testing these tools.

If your writing skills are not better than an AI’s writing skills, now is the time to either level up your writing skills or learn how to operate AI software effectively. There isn’t much middle ground on this – either you get better, or you work with the machines that are better. There isn’t a place at the table for mediocre to poorly skilled writers in the very near future.

At a societal level, we need to solve for some very important issues sooner rather than later, things like universal basic income. As I said, the output today is meh at best. It’s not going to stay that way. We’re already seeing some publications announcing more layoffs of writers as generative AI tools are adopted as cost-cutting measures. That’s going to accelerate. Something like universal basic income is essential to keeping the economy operational, because if you reduce the number of employed people by 40-60% – which is very possible as these tools advance – you will need to provide for them in some fashion.

Of all the AI technologies I’ve seen demonstrated in the last year, autonomous AI is the first one that legitimately unsettles me. Watching the tool running on my laptop screen, seeing how it thinks and reasons – it’s unnerving. As its quality improves, as it can tackle more complex tasks and more nuanced tasks, I believe it poses as many dangers as it does benefits, perhaps more. You owe it to yourself to get smart about it and watch it carefully as it evolves to see what the big picture implications are sooner rather than later. I know I am.

Got a Question? Hit Reply

I do actually read the replies.

Share With a Friend or Colleague

If you enjoy this newsletter and want to share it with a friend/colleague, please do. Send this URL to your friend/colleague:

https://www.christopherspenn.com/newsletter

ICYMI: In Case You Missed it

Besides the newly-refreshed Google Analytics 4 course I’m relentlessly promoting (sorry not sorry), I recommend the bakeoff that Katie and John did this week while I was traveling on generative AI tools, especially in the context of podcasting and video.

Skill Up With Classes

These are just a few of the classes I have available over at the Trust Insights website that you can take.

Premium

Free

Get Back to Work

Folks who post jobs in the free Analytics for Marketers Slack community may have those jobs shared here, too. If you’re looking for work, check out these five most recent open positions, and check out the Slack group for the comprehensive list.

Advertisement: LinkedIn For Job Seekers & Personal Branding

It’s kind of rough out there with new headlines every day announcing tens of thousands of layoffs. To help a little, I put together a new edition of the Trust Insights Power Up Your LinkedIn course, totally for free.

👉 Click/tap here to take the free course at Trust Insights Academy

What makes this course different? Here’s the thing about LinkedIn. Unlike other social networks, LinkedIn’s engineers regularly publish very technical papers about exactly how LinkedIn works. I read the papers, put all the clues together about the different algorithms that make LinkedIn work, and then create advice based on those technical clues. So I’m a lot more confident in suggestions about what works on LinkedIn because of that firsthand information than other social networks.

If you find it valuable, please share it with anyone who might need help tuning up their LinkedIn efforts for things like job hunting.

What I’m Reading: Your Stuff

Let’s look at the most interesting content from around the web on topics you care about, some of which you might have even written.

Social Media Marketing

Media and Content

SEO, Google, and Paid Media

Advertisement: Google Analytics 4 for Marketers (UPDATED)

I heard you loud and clear. On Slack, in surveys, at events, you’ve said you want one thing more than anything else: Google Analytics 4 training. I heard you, and I’ve got you covered. The new Trust Insights Google Analytics 4 For Marketers Course is the comprehensive training solution that will get you up to speed thoroughly in Google Analytics 4.

What makes this different than other training courses?

  • You’ll learn how Google Tag Manager and Google Data Studio form the essential companion pieces to Google Analytics 4, and how to use them all together
  • You’ll learn how marketers specifically should use Google Analytics 4, including the new Explore Hub with real world applications and use cases
  • You’ll learn how to determine if a migration was done correctly, and especially what things are likely to go wrong
  • You’ll even learn how to hire (or be hired) for Google Analytics 4 talent specifically, not just general Google Analytics
  • And finally, you’ll learn how to rearrange Google Analytics 4’s menus to be a lot more sensible because that bothers everyone

With more than 5 hours of content across 17 lessons, plus templates, spreadsheets, transcripts, and certificates of completion, you’ll master Google Analytics 4 in ways no other course can teach you.

If you already signed up for this course in the past, Chapter 8 on Google Analytics 4 configuration was JUST refreshed, so be sure to sign back in and take Chapter 8 again!

👉 Click/tap here to enroll today »

Tools, Machine Learning, and AI

Analytics, Stats, and Data Science

All Things IBM

Dealer’s Choice : Random Stuff

Advertisement: Ukraine 🇺🇦 Humanitarian Fund

If you’d like to support humanitarian efforts in Ukraine, the Ukrainian government has set up a special portal, United24, to help make contributing easy. The effort to free Ukraine from Russia’s illegal invasion needs our ongoing support.

👉 Donate today to the Ukraine Humanitarian Relief Fund »

How to Stay in Touch

Let’s make sure we’re connected in the places it suits you best. Here’s where you can find different content:

Events I’ll Be At

Here’s where I’m speaking and attending. Say hi if you’re at an event also:

  • B2B Ignite, Chicago, May 2023
  • ISBM, Chicago, September 2023
  • Content Marketing World, DC, September 2023
  • MarketingProfs B2B Forum, Boston, October 2023

Events marked with a physical location may become virtual if conditions and safety warrant it.

If you’re an event organizer, let me help your event shine. Visit my speaking page for more details.

Can’t be at an event? Stop by my private Slack group instead, Analytics for Marketers.

Required Disclosures

Events with links have purchased sponsorships in this newsletter and as a result, I receive direct financial compensation for promoting them.

Advertisements in this newsletter have paid to be promoted, and as a result, I receive direct financial compensation for promoting them.

My company, Trust Insights, maintains business partnerships with companies including, but not limited to, IBM, Cisco Systems, Amazon, Talkwalker, MarketingProfs, MarketMuse, Agorapulse, Hubspot, Informa, Demandbase, The Marketing AI Institute, and others. While links shared from partners are not explicit endorsements, nor do they directly financially benefit Trust Insights, a commercial relationship exists for which Trust Insights may receive indirect financial benefit, and thus I may receive indirect financial benefit from them as well.

Thank You

Thanks for subscribing and reading this far. I appreciate it. As always, thank you for your support, your attention, and your kindness.

See you next week,

Christopher S. Penn


You might also enjoy:


Want to read more like this from Christopher Penn? Get updates here:

subscribe to my newsletter here


AI for Marketers Book
Take my Generative AI for Marketers course!

Analytics for Marketers Discussion Group
Join my Analytics for Marketers Slack Group!