Hi everyone! I feel like I’m swimming out to try and body-surf the Ai wave, but fear that I’m too slow and the wave will pass, and I’ll be eaten by the sharks!
Do any of you have suggestions of a good place, or a class, to learn about Ai? I realize this field is moving faster than a one-legged man in a butt kicking contest, but I’d be happy to just know more than how to enter a basic prompt (question) into ChatGPT and get an answer. I feel like there is so much more I’m missing, so am hoping some of you have taken classes, or found really good tutorials, that you’d share with me. Thanks!
This is not bad:
Google offers some courses (but they also sell AI solutions, so the outlines might be better than the detailed ones, unless that’s what you want)
You might be overthinking it. Start just by doing something that you would usually use Google for. See what you get, ask a follow up question as if you were talking to a person. Ask it how to do something in an app you like. Then do a Google search for the same thing and compare the answers (and/or read the links it provided).
At that level it is little more than a web search though. It’s just giving you data it scraped from a site.
It gets complicated when you want it to use your own data as a source using your own LLM. If that is what you want, then there are numerous tech YouTubers with videos on it, but it is much more complicated. You can do it for free though.
It is easy to think you are “missing the wave” or you will be washed over. Don’t buy the hype. There are plenty of snake oil sales folks working to make you feel you’re missing out. Even Grammarly is now claiming to be a “trusted AI partner” or some such. Logitech added AI to their mouse driver! Everyone wants to sell you, or investors, something. Relax. You’ve plenty of time to catch up.
There are valid use cases for these tools. Folks are using them to help with writing, help with coding, help with brainstorming, help with image generation. If there are things you do that fall into such categories then look for introductory courses on those topics. Note that it helps to have some competence in the area you wish to use the tool in. Just like using power tools won’t make you a contractor, using LLMs won’t make you a (good) writer or programmer. These tools are best used to enhance one’s abilities, not provide new ones (but they can help you learn).
Finally, trust but verify. As mentioned, the capabilities of these tools are often overhyped. There have been several examples on this forum where the answers from an LLM are just straight up BS. A couple of viral examples such as using glue to hold cheese on pizza or getting one’s daily doses of minerals by eating a small rock, are easy to spot. Others are more subtle, but always be aware the LLM has no understanding. It is all just words strung together. A recent example I saw was where a LLM was asked to provide a review of the movie Wicked. And it provided a very reasonable review. Cool, right? But realize no LLM has ever seen a movie. It is all made up. Unlike with a human collaborator, LLMs don’t ‘know’ what they don’t “know”, but they will confidently provide an answer anyway. This is a key insight to understand when using these tools.
Getting back to the initial question, if you are just interested in a general course, the Google one mentioned by @chrisecurtis seems as good as any. If there are specific areas you are interested in let us know and perhaps we can provide more specific options.
And as @RunningBoris said, just start playing with them. I’m having fun using LLMs are a coding companion learning Python. They can be both dumbfounding and delightful in the results they provide.
Don’t worry, the world has barely dipped its toes in the AI wave. In a way this reminds me of the early days of search engines. In the early 90s I had recently set up our first LAN and been connected to the internet (at 56K) when I was asked if I could get a list of multiple religious holidays for the personnel department.
I went to (I think) Alta Vista got the information and gave it to them a few minutes later. They were amazed. A few weeks later everyone knew how to search for information.
All of us won’t need to be AI experts. Today we are dealing with components of what will become agents. I won’t need to know much when I can say “Siri, book me a flight to Denver for next Tuesday. And I’ll need a room at the Hilton, and Ubers to take me to and from the airport.”
Today that is the future but we are seeing early examples of what an AI agent can do.
Introducing Gemini 2.0: our new AI model for the agentic era
I am wading through a lot of dreck—doubtless due to my lack of knowledge in this area. Any specific videos or YouTubers that you recommend to begin here? This would be tremendous for my work.
I was watching NetworkChuck’s tutorial, but he is a bit much at times, and I am not sure I recommend it. I was following along, got it set up, but then really never did anything with it.
Ollama is the one that everyone recommends, so you can do a search for “Ollama tutorial” and get a lot of hits.
If you are interested in Ollama, which is what I’m currently using, then you might find these videos helpful:
Specifically his Introduction to Ollama playlist.
Have fun!
+1
@RickPagels, if you continue to follow the MPU forum, you won’t miss anything big in AI. And you can learn along the way. Don’t drink the Kool-Aid.
The way I look at all this AI:
Text tools (aka large language models - LLM) - Help with ideas, writing, outlining, summarizing, and research (can even replace traditional Google or other Search engine for searches).
Image (and now video) generation - text based prompted or newer tools that take existing image or videos as examples/templates to augment.
AI-driven features inside other apps (not just convenient pass-thru to ChatGPT or equivalent, but true use of AI like voice processing tools inside audio editors to remove background noise, improve audio quality, etc. with ‘one-click’ use)
Resources to learn:
#1 - try and experiment. Many tools are free or have free trials.
#2 - YouTube videos - lots of free stuff for training, usage, or news to stay current on what is available and how to use them
#3 - Blogs, books, ebooks - Lots of free and some paid, but I prefer video and other formats where they can demonstrate with screen videos, not just textual descriptions or static screen shots.
#4 - Paid courses - more in-depth, directly from authors or through online training platform websites. Some cost per class, some part of flat-rate monthly membership fees.
Best way I learn:
Start with a goal instead of random trial and error.
Make it small and quick so easy to iterate quickly.
Ideas:
Take existing writing and run it thru AI with prompts to “make it more concise”, “make it more exciting”, “make it easier to understand by someone non-technical”. ChatGPT, Gemini, or Claud are the big tools to try.
Create an image to illustrate something you have or already making. A blog post, a marketing flyer, an article or newsletter, or an email. Web-based Adobe Firefly is a good start.
Research a topic of interest or something you are working on and try it both using AI and traditional search engine or web site research and compare the results. Google NotebookML is great for this.
I find ChatGPT 4.o to be extremely helpful. I ask AI to “give me a point x miles from the start of my trip, give me a list of hotels with star ratings of no less than 4 in a price range of x. Give me the address, phone number webpage and price. Provide the information in a markdown file for downloading.” It works very well for this sort of thing–must faster than googling for the information.
How’s the accuracy for this sort of thing?
Do you hop in the car or board an airplane without doing some heavy-duty confirmation of what your useful friend provided?
I found it to be accurate now that ChatGPT can search the internet. I always verify, which I also have always done with a standard Google search.
Wouldn’t it be easier to just skip the middle man then and just use Google?
ChatGTP uses something like 5 times the energy of a Google search. AI/Crypto is going to put a huge strain on our resources.
For a great introduction to the uses and problems with AI, the best resource I am aware of is the book, Co-Intelligence: Living and Working with AI by Ethan Mollick.
No - ChatGPT for now is pure text. No sponsored results, no ads, no pollution.
Sure, it like early Google before they decided they needed a revenue stream to stay around,
So enjoy a throwback, yet better, way to search the web until AI stops being a fad and VC race and folds into the ensh**tification of the Internet that is inevitable for AI too?
Except you can’t trust it. You get the popular answer, not necessarily the correct or best answer. It, and other AIs, have presented me with incorrect info plenty of times. I also prefer to pick my source. If I have an Excel question I know which sites I trust. The person I was asking, said they often verify the answer, which means you were better off using Google in the first place.
Also, the sites that have good info deserve credit and the traffic.
Then you have the ridiculous power requirements…
This is what brings me back to Perplexity for many searches. Yes, results for important queries still require verification, but I have yet to be burned.
Yes>>
- I’m taking this Coursera class, it’s just what I wanted for a foundation, overview, defining of terms like LLM / Machine Learning / Neural Net:
AI For Everyone
from Deep Learning.ai / Andrew Ng.
How did I hear of it? by asking ChatGPT how I could learn more about AI! Next I might do the VanderbiltU AI ones.
- How I learned to use ChatGPT? these two episodes of ScreencastsOnline: