The 3-Minute Guide to Slaying Your Dinner Convo About AI
Welcome to your cheat sheet on AI. These talking points will help you survive and thrive in almost any conversation on AI (assuming you’re not in a room full of MIT researchers, of course!)
The 1-liner summary of what’s going on in AI
AI models are getting advanced very fast, so there’s a rush by companies and professionals to figure out where they can be used, along with grappling with some big questions about what roles AI should or shouldn’t play in society.
The 1-paragraph summary of what’s going on in AI
We're in the thick of an AI revolution. Over the past decade, tech companies and research firms have pushed the pedal to the metal, developing models (i.e., algorithms) that now astound us with text, images, and sound that appear to be made by us humans. It’s … a lot to take in. From pondering how AI-powered tools will affect our daily routines to the larger existential questions (i.e., will AGI kill us?), we're wrestling with it all. In business, AI is like a new slide on the playground - everyone wants a piece of the action. Meanwhile, regulators are scrambling to set rules for this new world. You and me, on the other hand? We’re just trying to make sense of it all, and that starts by understanding exactly how this thing called “AI” works.
Flex your knowledge on ‘em
- AI (Artificial Intelligence): AI is like a clever robot that can understand, learn, and make decisions. It's a broad term covering any software designed to mimic human intelligence.
- Machine Learning: An AI improving over time by learning from data.
- Language Model: Given a bunch of data, and language model (LM) can predict the next word or finish your sentence.
- GPT (Generative Pretrained Transformer): GPT is a language model trained on loads of text data to generate human-like responses. Yup, this is how ChatGPT works.
- Reinforcement Learning: It's like training a dog. You tell an AI if it’s good or bad, and it learns to improve over time.
- Prompt Engineering: This is like carefully phrasing a question (or prompt) to get the best possible answer. It’s more prompting and less engineering.
- Deepfakes: Deepfakes are AI-generated images, videos, or audio files that convincingly mimic real people. Like those images you saw of The Pope.
- AGI (Artificial General Intelligence): A superintelligent AI that can handle any intellectual task. Tech nerds see this as AI's final state, which either brings about utopia on Earth or kills us all…
Fun tools that you can bring up
ChatGPT - ask an AI chatbot anything
Midjourney / Stable Diffusion / Adobe Firefly - generate images
Rewind - remember everything you see on your Mac
Tome - create slide decks
Descript - all-in-one video/audio editing
…plus tools like Jasper to help you write anything: blog posts, summaries, cover letters, resumes, emails, and essays.
The names you should know
OpenAI - They produced leading models GPT and DALL-E, an AI that beat champions in DotA 2, and the AI that powers Github’s AI coding tool. Their mission is to develop safe artificial intelligence … and, ultimately, AGI.
NVIDIA - US-based chipmaker that produces the computer hardware researchers use to train AI models. By far the leader in this space, and its stock certainly reflects that.
Microsoft - Invested $1 billion in OpenAI and helps OpenAI train their models. They’re integrating AI everywhere in their Office ecosystem and even in their operating system, Windows 11.
Google - Is working on multiple competing models to OpenAI’s GPT and DALL-E. The biggest question is how Google monetizes and upends search in the ChatGPT era (they’re experimenting with AI search hard)!
DeepMind - Google subsidiary and AI research company that became popular after it created an AI so good at the board game Go that it pushed the top player to retire.
Sam Altman, Sundar Pichai, Satya Nadella - OpenAI CEO, Google CEO, Microsoft CEO
The EU AI Act - The first major piece of legislation attempting to regulate AI companies and models (many in tech think it’s too harsh).
What’s different about this wave of AI models?
Gone are the days when AI models only mastered single tasks like recognizing handwritten numbers. Today's models, known as foundational models, are versatile powerhouses trained on a broad spectrum of data. But their adaptability doesn't stop there. They can also be fine-tuned for specific tasks, offering the best of both worlds. And there's more. With the arrival of open-source foundational models like Meta's LLaMA, the development world is thriving.
What’s the debate about AI art?
The controversy comes in a few parts:
- AI models need data to learn. They take this data from human artists’ work. But how do we know whose artwork inspired the model’s output, and what do we owe them, if anything?
- Then there’s a larger existential question: Does the ability to so easily create any art of any style ruin the spirit of art itself?
You can imagine that an artist and a technologist would have very different answers about these!
Is AI good or bad for society?
It largely depends on what you’re using AI for. People believe automating low-level work is a good thing. But things like knowledge work, making decisions, and telling humans what to do are controversial. That’s because we aren’t confident that we can always get AI to do what we want it to do.
Personally, we think the good AI harnesses will vastly outweigh the bad. But hey, we’re optimists!
How private is AI…?
As we increasingly chat with AI, the question of 'how much stays just between us?' looms large. For the most part, the secrets you share with ChatGPT stay just that - secrets. However, these chats do contribute to the model's training, raising some eyebrows in the privacy-conscious corners. Many companies are pulling on the brakes, barring ChatGPT in the workplace due to data privacy and security concerns. This has sparked a surge in efforts by enterprise SaaS companies to develop AI cloud solutions that ensure your internal data stays internal.
Are we going to lose our jobs because of AI? When will that happen?
AI will certainly make some jobs obsolete, but it’ll also create new ones, just as every major technological shift has. One thing is for sure: we probably need to reskill the workforce in light of AI, but exactly who and when are big questions. If there are massive shifts in the economy because of AI, we’ll also have to think about ensuring people can continue to earn their living.
Is AI going to become sentient / develop consciousness / have emotions?
What is consciousness anyways? How do you measure that?
AI will eventually be able to pass any test that you design. For example, if the question is “How are you feeling?” and the AI responds with “Sad,” can we conclude that it has feelings?
And will it take over the world, like Skynet in Terminator? Probably not, but that will be up to us to decide what power we give it.
Phewww! That was a lot to cover for one day, but congrats - you now know more about AI than 99% of humans.
If you want to become an expert in AI and truly impress your coworkers, check out our brand-new website with highly vetted “AI Tools For Business” and tutorials on how to use them (coming soon).
How else can AI help you? Let us know what types of tutorials you want to see from us next here!