Getting Started with AI

With all the hype surrounding artificial intelligence (AI), it can be intimidating to figure out how to get started. 

What is AI? What’s the big deal with ChatGPT? What does GPT stand for? How should I use AI? Which AI tools should I use? Should I pay for AI? 

These are questions that come up in every workshop, training session, and conversation I have with folks just getting started in AI.

Let me help you answer these questions and more. 

Brief history of AI

Artificial intelligence is a field that has been around since the 1950s. It generally refers to the concept of computer systems that are capable of human reasoning, but even computer scientists don’t all agree on the definition of AI. It encompasses many different tools and technologies including predictive AI, generative AI, machine learning (ML), deep learning (DL), natural language processing (NLP), natural language understanding (NLU), sentiment analysis, computer vision, large language models (LLMs), robotics, neural networks, and more. 

You’ve been using products and services that use AI — and you’ve been using them for decades. All of the recommendation engines — Amazon, Netflix, Spotify — use ML for predictive AI. Alexa, Siri, and Google Assistant all use AI technologies NLP and NLU to understand speech. Even Gmail, Google Docs, or Grammarly helping you complete a sentence is AI. 

Generative AI

Earlier types of AI were predictive, meaning they  analyzed existing data to make predictions about future outcomes.

Generative AI creates new content based on patterns learned from training data. When ChatGPT was released in November 2022, it was the first publicly available generative AI for conversational use. This was groundbreaking because it made AI accessible to the masses in an interactive, conversational format. It showcased the ability of AI to generate human-like text on demand, opening up a world of possibilities for how AI can assist with content creation, problem-solving, analysis and more.

But generative AI is only as good as the data it’s trained on. So if it’s been trained on biased data, you’ll get biased output. And these AI systems are hungry for data (which is why they have free versions to get you to give it data), so it’s best to assume that anything you type into it could be used as training data and could one day become public. 

What does GPT stand for?

In ChatGPT, GPT stands for “Generative Pre-trained Transformer.” ChatGPT tells me, “This refers to a class of language models developed by OpenAI that are trained to generate human-like text based on input prompts. They use transformer architecture and are pre-trained on large datasets to understand and generate text.”

Ironically, GPT also stands for “General Purpose Technology.” GPTs are versatile technologies with broad applicability across various sectors that leading to significant productivity enhancements and transformative societal changes. GPTs have historically shaped economies and societies, driving progress and innovation.

Examples include electricity, which revolutionized manufacturing and communication; the steam engine, enabling mechanized production and transportation during the Industrial Revolution; computers, which have changed how we process data; and the internet, connecting people globally and facilitating information exchange. We can be pretty certain that AI is also a GPT that will revolutionize industries through automation and intelligent systems. 

Frameworks for thinking about AI

If you’re just getting started with AI, these frameworks may help: 

• Helpful intern: AI can take on tedious research and drafting tasks, freeing up your time for higher-level work. But like an intern, it needs context and lots of guidance and oversight. And just like you wouldn’t fire an intern if it can’t do things perfectly right away, you’ll have to be patient with AI and train it to understand what you need. 
• Thesaurus for whole thoughts: AI can suggest alternative phrasings to refine your writing and convey your points more effectively. This is how I personally use AI most frequently. 
• Personalized search engine: AI can quickly find relevant information and insights tailored to your specific queries, without you having to sift through pages and pages of search results. .  
• Editor: AI can proofread your work, offer editorial suggestions, and help ensure consistency of voice and style. I have more tips on using AI for rewriting here. 
• Teacher/coach: AI can explain complex topics, provide practice problems, and offer feedback to accelerate learning. I love using AI to help my children get additional practice problems and to help me learn something I’m confused about. “Can you help me understand what this block of code is doing?” is something I ask ChatGPT regularly. 
• Idea Generation / Brainstorming Buddy: AI can help generate novel ideas, explore “what if” scenarios, and inspire creative thinking. I also love it for brainstorming brand names or company names.

Underscoring all of these frameworks is that AI today is a co-intelligence. It’s best used to augment or enhance your existing capabilities. 

Because it was trained on human data, it responds the way humans do. Be polite and it behaves better than you berate it. But like humans, it is fallible, prone to making up stuff (“hallucinations” in AI-speak), unpredictable, and full of all of our human biases. And yet, as of May 2024, it is not human (one day we’ll get to artificial general intelligence but that’s a topic for another day). 

I like to think of it this way —  every single one of us now has immediate 24/7 access to personal assistants. 

Those who don’t know how to use AI will be replaced by those who embrace it. 

AI Tools

If you’re just getting started, you’re likely wondering which tools to start using. The complicated answer is to use them all. These tools are improving rapidly and new tools are being released that its impossible to find an up to date guide for which tools are best at that moment. 

The best way to discern what’s best for your use case at that moment is to try it out for yourself. Generative AI use is still so new that if you dive in to how it can solve your specific problem, you could very well become the expert on using generative AI for that use case. 

That said, these are the basic tools I’d start with for generating text, images, songs, presentation decks, text-to-speech, and even text-to-video. 

Text Generation (LLMs):

Each different large language model has its pros and cons. I don’t want to go deep into them but I’ll briefly describe each of the main LLMs. 

• OpenAI’s ChatGPT: tends to be a bit too flowery and robotic but seems to be the best with code.
• Anthropic’s Claude: has the best privacy policies and can be HIPAA compliant.
• Meta’s Llama: as of May 1, 2024, this open source LLM seems to be better than free ChatGPT and it’s incorporated into every Meta product including Facebook and Instagram. Now that OpenAI is releasing ChatGPT 4o to all free users, ChatGPT will once again be the best free LLM.
• Perplexity: cites its sources and functions mostly as a better search engine (all the other LLMs will make up sources).
• Inflection’s Pi: this one is the most conversational and is built to be supportive and empathetic.
• Microsoft Copilot: built in to every Windows machine and is built on ChatGPT
• Google Gemini: has a more natural, conversational tone and is linked up to most Google services. 

For your specific use cases, you’ll want to try several of the models to figure out which one works best for you.

Personally, I use a mix of all of these — yes, all of them — and others.   

Image Generation:

Today’s image generation tools use what we call diffusion techniques and are improving rapidly. But they often still generate extra body parts (especially limbs and fingers) and weird text that is almost, but not quite, correct. It’s always best to carefully review a generated image before using it. 

• Adobe Firefly: currently free (as of May 2024) and designed for commercial use as it’s been trained mostly on Adobe-owned images. 
• Canva: incorporates several different AI image tools in its Magic Studio. 
• Ideogram: 100 free images every day. 
• Midjourney: must use Discord to create images. Not sure what Discord is? Move on to another tool. 
• Stable Diffusion: released in August 2022 before ChatGPT came on the scene. 
• DALL-E: included in the paid version of ChatGPT. 

Generative AI for Songs:

Creating your own songs in a specific style on a unique topic is genuinely delightful. Reggae, hard rock, jazz — it can do it all. 

• Suno: more melodic and better lyrics. I have a friend who now prefers suno music to human-created music. 
• Udio: more customizable and better with chord progressions. 
• AI Jukebox: easy to use with ability to set the duration of the song.

Presentation decks:

If you’re like me and have given lots of presentations, you’ve always dreamed of having your PowerPoint presentations made for you. Canva has certainly made it easier to make gorgeous presentation slides, but you still have to make them. The next generation of tools takes it to the next level:

• Beautiful.ai: easy-to-use presentation software with smart templates for quick, professional designs. As of May 2024, no free tier. 
• Gamma: instantly personalize decks and webpages with AI, ideal for GTM (go-to-market) teams.
• Tome: converts text into visually appealing slides, supports 100+ languages, geared towards sales and marketing professionals. 
• SlidesAI: quickly generates professional slides from any text, integrates with Google Slides.

Text-to-speech:

• ElevenLabs: text to speech in any voice, language, and style. 
• Resemble: clone your own voice.

Meeting Assistants:

Most meeting assistant tools use AI to create automated summaries and action items. Here are a few options:

• Otter: transcripts, automated summaries, action items, and chat to get answers from your meetings.
• Fathom: AI notetaker known for its compliance requirements
• Fireflies: automated meeting notes, action items, and transcripts that can add entries in your CRM.
• Sonix: transcribes both audio and visual context into text.
• Circleback: AI-powered notes, action items, and automation; automatically updates your CRM, Notion, and more.
• Krisp: AI-powered noise cancellation, transcriptions, meeting notes and recording.

Workflow Automation:

• Zapier: not technically an AI tool, but this no-code workflow automation tool allows anyone to connect web apps and services to automate things like social media, customer service, project management, and more.
• Tonkean: powered by AI, Tonkean helps enterprise internal service teams like procurement and legal create process experiences that people actually follow.

Scheduling Automation:

• Clockwise: AI scheduling and calendar automation.
• Reclaim AI scheduling to defend focus time, automate meeting scheduling, provide time tracking analytics, and improve work-life balance.
• Motion: AI scheduling assistant to manage projects, tasks, and calendars.

Text-to-video:

• Canva: turn photos or videos into talking heads for any project. 
• Synthesia: no free tier but they do have a free demo. 
• Luma’s Dream Machine: free version requires sign-in with a google account
• Runway’s Gen-3: easy to use, with a free version

The best Photorealistic tools like OpenAI’s Sora and Irreverent Labs aren’t publicly available, yet. Google Deepmind’s Veo has a waitlist open as of May 17, 2024. 

Again, these lists are by no means comprehensive. They are a starting point. There are also lots of niche products that might work very well for specific uses. These tools are changing every day and once you gain some confidence you’ll want to continue exploring and experimenting with all the latest technologies and advancements. What doesn’t work in one AI tool today may work very well in it next month. There is no handbook. 

I’ve been giving talks on the future of work since 2020, experimented with generative AI since Stable Diffusion and ChatGPT came on to the scene in 2022, and I’ve taken graduate-level coursework in artificial intelligence, neural networks, natural language processing, and generative AI. But even I don’t call myself an expert because the technologies and tools keep changing. Experiment and iterate and become your own expert.

Free versus paid

Almost every single AI tool has a paid version and a free version. As I mentioned above, the free versions exist so these companies can get more data — from you! The free versions are generally good enough, but the paid versions are significantly better. If you can’t afford it, the free versions will be just fine. But if you can, I’d pay up for paid versions of the top 3 LLMs of today — ChatGPT, Gemini, and Claude. 

Tips for using AI

Finally, I’ll end with a few quick tips for using AI. 

• Iterate: If an Al tool’s output isn’t perfect, work with it with different inputs or settings. Have a conversation to help it understand what you need. 
• Verify: Always double check information generated by AI. Triple check it if it’s important. Trust, but verify, as my favorite Russian proverb warns.
• Share wisely: Be mindful about privacy settings and the information you share with AI tools. I personally don’t share anything with AI that I wouldn’t be comfortable becoming public. 
• Experiment: Try out lots of different AI tools to find the ones that best fit your needs. Just because AI can’t do it today doesn’t mean it won’t be able to do it next week. 
• Human-in-the-loop: Use AI tools as an assistant, not a replacement for human interaction and thought. 

If you’re interested in going deeper, I share more tips on prompt engineering here. 

Start exploring AI 

Diving into AI opens up a world of possibilities. This powerful tool is accessible to everyone. Use it as your own personal assistant for anything you currently find tedious or that usually requires another human being — practice for an interview, create calendar event import files, and much more. Stay curious and adaptable as AI continues to evolve. The future belongs to those who can effectively work with AI.

So dive in and start exploring!