Category: Artificial Intelligence

OpenAI needs bodyguards…er, “Executive Protection Operators”

Post author By Joey deVilla
Post date March 4, 2024

While perusing OpenAI’s “Careers” page, I noticed that the drop-down menu that lets you filter jobs by department included Corporate Security, and that doesn’t mean cybersecurity, but security of much rougher, tougher, old-fashioned sort.

At the time of writing, this department has just one job listing, and it’s not one you’d expect to see for a software company:

Executive Protection Operator

About the team
The Corporate Security (CorpSec) team at OpenAI is dedicated to ensuring the safety and security of our people and facilities. We are committed to maintaining a secure environment that enables our team to focus on advancing artificial intelligence in a responsible manner.

About the role
As an Executive Protection Operator, you will play a crucial role in safeguarding the well-being of OpenAI’s executives and key employees. Your responsibilities will encompass providing security support during travel, events, and, when necessary, day-to-day operations. You will need to have proficiency in firearms handling and the legal authority to carry a firearm in the United States. You are a seasoned, collaborative security professional with a deep understanding of executive protection principles and a proven ability to navigate complex, dynamic security situations with discretion and tact.

This position is based in our San Francisco HQ. We use a hybrid work model of 3 days in the office per week, though this role is expected to have a flexible schedule that aligns with the travel and operational needs of the executives you support. We offer relocation assistance to new employees.

In other words: they’re looking for bodyguards.

Clearly the folks at OpenAI — not normally the best at “reading the room” — understand that the “creative destruction” that their products will unleash upon the world over the next few years will make them deeply unpopular in certain circles. It appears that they’re taking precautions (and hey, it might have been a suggestion from ChatGPT!).

My favorite part of the posting is the “desired qualities” list, especially the last item:

You might thrive in this role if you:

Hold a current and valid Law Enforcement Officers Safety Act (LEOSA) card for concealed carry in all 50 states, with 10+ years of related work experience or at least 5 years with an advanced degree.

Maintain the highest level of confidentiality and discretion, with a proven ability to stay composed in urgent and high-pressure situations.

Have experience in executive protection for top-level leaders in diverse environments and have completed recognized Executive Protection and driving courses.

Possess leadership experience coordinating with domestic and international law enforcement, military, intelligence, and corporate partners.

Are trained in tactical medical procedures and have a strong understanding of Google Workspace applications.

That last bullet point might be the first time the qualifications of tactical medical procedures and strong understanding of Google Workspace applications have been combined in a job requirement. Applicants might do well to write this in their cover letter:

I’m a medic in the streets, and a beast on Google Sheets!
No need to credit me with this line.

Are you qualified? Are you willing to take a bullet for Sam Altman or an equally machiavellian AI bro? Apply here.

And because the song from The Bodyguard is now stuck in my head, I’m sticking it in yours, too:

Artificial Intelligence Humor

When you ask ChatGPT for a word that begins with “i”

Post author By Joey deVilla
Post date March 3, 2024

Screenshot of ChatGPT conversation:

Me: I need a word that begins with "i" that is a synonym for speech or presentation

ChatGPT: A word that begins with "i" and is a synonym for speech or presentation is "address."

Me: "Address" doesn't begin with "i."

ChatGPT: Apologies for the oversight. A word that begins with "i" and is a synonym for speech or presentation is "oration."

Me: That doesn't begin with "i" either.

ChatGPT: You're right, my apologies for the oversight. A word that begins with "i" and is a synonym for speech or presentation is "interlocution."

Me: There you go. Nicely done.

Here’s your daily reminder that large language models don’t actually “understand” the world — at least not in the same way that we do. They’re stochastic parrots.

Tags ChatGPT, LLM

Artificial Intelligence Conferences What I’m Up To

I’m speaking at Civo Navigate North America in Austin next week!

Post author By Joey deVilla
Post date February 13, 2024

In exactly one week, I’m going to giving my presentation, You’re Not Too Late to the AI Party at the Civo Navigate North America conference in Austin, Texas!

The talk, which is in a prime time slot — day one (Tuesday, February 20) just after the opening keynote (10:30 a.m.) — is for people who’ve been too busy with their actual work to get into AI and have been feeling increasing amounts of FOMO.

I’ve been spending my suddenly copious amounts of free time polishing this presentation and accompanying demos to a bright sheen. Here’s a sample from my current set of slides:

In the talk, I’ll cover four reasons why it’s not too late to get into AI, as well as possible ways you can get started.

I’ll also talk about:

AI’s overnight success was 70 years in the making
Are you a centaur or a minotaur, and which one is better?
AI options if you can code, and options if you can’t
AI techbros and “chaos muppets,” and the effect they have on the industry
AI ethics and how badly we need it

This talk won’t be all hand-wavey and descriptions, but will also feature demos of actual working code that you can also download, including:

ELIZA, the original 1964 chatbot, but written in present-day Python.
A basic neural network demo that shows how you implement them — perhaps the one that recognizes handwritten numbers, perhaps something a little more interesting!
“Sweater or no?” — a large language model-powered application that tells you what to wear based on your location, the weather, and the event you’re attending.

Most importantly, the talk will be fun!

I’ll be in Austin for most of next week. If you can attend Civo Navigate, I’d love to see you there! I’ll also be free and out and about in Austin that Thursday (Feb 22) for most of the day — if you’d like to meet up, let me know!

Tags Civo, Civo Navigate

Artificial Intelligence Geek Culture

The first AGI meme of 2024: “Look what they need to mimic a fraction of our power!”

Post author By Joey deVilla
Post date January 3, 2024

In response to my reposting of Buzz Andersen’s spot-on post on AGIs, Tampa Bay area student Jason Lee made a comment that had me laughing out loud — “Look what they need to mimic a fraction of our power!”

In case you don’t get the reference, it’s a line from the animated version of Invincible, the streaming adaptation of the comic book with the same name. Here’s a clip from the episode where it’s uttered:

So for now, think of us humans as the Viltrumites and AIs as puny humans.

For now.

Artificial Intelligence

The best social media post about AGI that you might have missed

Post author By Joey deVilla
Post date January 2, 2024
2 Comments on The best social media post about AGI that you might have missed

Buzz Andersen’s recent post on Threads about AGI — artificial general intelligence — got a lot of attention, even though he posted it during the slowest week of the year:

Buzz Andersen’s (@ldandersen) post on Threads: “I was talking to one of my old co-workers who works in machine learning for a big tech company awhile back, and when the subject of “AGI” came up, he said something like (and I’m paraphrasing here): “These models require massive infrastructure, enormous amounts of power, and basically the entire Internet as training data. Meanwhile, the human brain learns from the world around it and runs on sandwiches.” I think about that a lot.” — Tap to view the original post on Threads.

Here’s the text:

I was talking to one of my old co-workers who works in machine learning for a big tech company awhile back, and when the subject of “AGI” came up, he said something like (and I’m paraphrasing here): “These models require massive infrastructure, enormous amounts of power, and basically the entire Internet as training data. Meanwhile, the human brain learns from the world around it and runs on sandwiches.” I think about that a lot.

Food for though (pun intended).

Artificial Intelligence Conferences Programming What I’m Up To

I’m doing an AI presentation at Civo Navigate North America in Austin in February!

Post author By Joey deVilla
Post date January 1, 2024

Conference talk title card for Civo Navigate North America 2024: “You’re not too late to the A.I. party” with Joey de Villa

It’s only day one of the new year and I just fulfilled one of my resolutions: to land a conference speaking session on AI outside my usual stomping grounds. I’m going to be a speaker at Civo Navigate North America, which takes place on February 20th and 21st in Austin, Texas!

What’s Civo Navigate, and what is Civo?

What’s Civo Navigate, you ask? Here’s a one-minute video that answers your question:

Civo is a cloud hosting provider based on Kubernetes, with a focus on developer-friendliness and wallet-friendliness. It’s a refreshing change from this state of affairs:

Comic about AWS

Panel 1: Person looking at venus flytrap, asking “Why would a fly land on something like this?”

Panel 2: Same person, looking at mousetrap: “Rats should be ashamed for falling into this trap.”

Panel 3: Same person, gesturing with both arms towards a bear leg trap: “BEARS! This is ridiculous!”

Panel 4: Same person looking at a sign that reads “AWS free tier:” “This must be fun!”

I met the people at Civo last year when they held Civo Navigate North America in Tampa — and not in a convention center or hotel conference rooms, but at Tampa’s big riverside food hall, Armature Works! Here’s the promo for that event:

The 2023 edition of Civo Navigate North America was a great conference with interesting talks and a warmer, more personal “feel” than a typical vendor-hosted event. Civo’s contributions continued long afterward, with their being great supporters of the Tampa Bay tech scene and this blog.

I’m looking forward to the 2024 edition in Austin?

What’s my talk about?

Title card: “You’re not too late to the A.I. party,” featuring a Canva AI-generated image of a party with two women in summer cocktail dresses and a robot at the bar. The generated image is deep in the “uncanny valley.”

My talk is titled You’re not too late to the A.I. party, and it’s for people who’ve been too busy with their actual work to get into AI and have been feeling increasing amounts of FOMO.

Here’s the description of the talk, with additional AI-generated photos (that are deep in the uncanny valley):

Have you been too busy getting your actual work done to join the artificial intelligence party and feel that you’ve already missed out on the technical career opportunity of a lifetime? If you answered “yes,” this talk is for you.

The good news is that you’re not too late to the A.I. party. It’s just getting started and you arrived at a good time — perhaps even “fashionably late!” You just need someone to take you around the room and make some introductions.

To help you “work the room” as you enter the party, you’ll get an overview of artificial intelligence technologies, from the rules-based models and expert systems of A.I.’s early days to the present era of neural networks, machine learning, transformers, and large language models.

This party won’t be limited to just hand-waving small talk in the living room. We’ll go into the kitchen — the true heart of any party — and look at actual code in action. We’ll start with ELIZA, the original chatbot from the 1960s, observe a neural network, and look at an LLM-powered “What should I wear today?” app. You’ll even be able to download them for yourself!

This talk aims to be like the best parties — the ones you’re glad you were at. You’ll leave this one knowing more about AI’s underpinnings and a much better idea of the next steps in your AI journey, whether it’s catching up with AI developments, harnessing your current skills to integrate AI into your work, or even pivoting into AI development.

In my talk, I’ll discuss:

Generative vs discriminative AI
“Old School” rules-based AI vs. the “New School” version powered by neural networks, data science, and lots of data
How the internet changed AI
The intersection of data science, statistics, and AI
The paper “Attention is All You Need,” what it means, and how it changed AI forever
Large language models (LLMs)
Retrieval-Augmented Generation (RAG)
Vector databases

This talk won’t be all hand-wavey and descriptions, but will also feature demos of actual working code that you can also download, including:

ELIZA, the original 1964 chatbot, but written in present-day Python.
A basic neural network demo that shows how you implement them — perhaps the one that recognizes handwritten numbers, perhaps something a little more interesting!
“Sweater or no?” — a large language model-powered application that tells you what to wear based on your location, the weather, and the event you’re attending.

Cover of the book “You Look Like a Thing and I Love You.”

I’ll also talk about potential “next steps” that you can take, including:

Reading material, including the funniest book about AI (for now): Janelle Shane’s You Look Like a Thing and I Love You. Of course, you don’t have to wait for the talk (or even attend) to read it; you can get it now!
There Will Be Math — or, the math you’ll need to know to get into AI.
Effective Altruists, Effective Accelerationists, and how to Effectively Avoid both.
How to send the right signals to employers so they’ll know that AI is your jam!

Find out more about Civo Navigate North America

Want to know more about Civo Navigate North America, my talk, and everyone else who’s presenting? Check out their site, and come on down to Austin for February 20th and 21st!

Artificial Intelligence Current Events Editorial

Analytics Insight’s very clumsy plagiarism

Post author By Joey deVilla
Post date November 30, 2023

On November 29, 2023, Analytics Insight — whose business model is pretty much “clickbait as a service” — published an article titled Ten Women AI Leaders to Enhance OpenAI’s Board, pictured in the screenshot below (along with a big error that I’ll point out shortly):

It bore a striking similarity to an article published three days earlier: 10 Women AI Leaders Who OpenAI Should Consider For Their Board, by Rebekah Bastian (SVP of Product at Glowforge), pictured in the screenshot below:

Later versions of Analytics Insight’s article included acknowledgment of the original Forbes article:

In spite of Analytics Insight giving credit where credit is due, their list of top ten women is markedly different. Here’s Rebekah Bastian’s list from the original Forbes article…

…and here’s the Analytics Insight list:

Fei-Fei Li
Rumman Chowdhury
Timnit Gebru
Mona Sloane
Joy Buolamwini
Yoshua Bengio
Kai-Fu Lee
Hinda Haned
Danielle Belgrave
Maja Pantic

Note entries 6 and 7, Yoshua Bengio and Kai-Fu Lee — neither of them are women, and their blurb about Bengio even uses a male pronoun:

6. Yoshua Bengio:

Deep Learning Pioneer: Bengio’s pioneering work in deep learning has had a profound impact on the AI landscape.
Educational Leadership: As a professor at the University of Montreal, he contributes to shaping the next generation of AI researchers and practitioners.

Consider this a timely reminder on the first anniversary of ChatGPT’s release (which happened November 30, 2022): For the moment, you still have to double-check an AI’s output before you publish it as your own!