The Usefulness of AI, and Tools That I Use

AI is a bit of a polarizing topic as of late. Most people fall into one of a few seemingly extreme camps. Either AI is the most revolutionary technology ever, and will trigger its own industrial revolution size change to our society, or it’s equally powerful but will cause more of an apocalypse of some sort, or it’s all just hype with no substance and seemingly completely useless. I don’t really think any of these are accurate, and it makes me sad that so many people have seemingly lost the ability to think or articulate nuance in complex topics. I think this is something that has been hyper-accelerated by social media, and is a huge problem. No topic is actually completely polar, but that’s how many people seem to talk about it any more.

I’m not going to pretend that I know where AI is going to go, or what it will be capable of in the future. I’ve heard people who are much smarter than me in the area make claims that Artificial General Intelligence (AGI, which would essentially be a “conscious” AI that could improve itself, much like humans, though there is some debate on the definition as well), is just a few years away, and others say things like the recently super popular Large Language Models are very impressive but are likely to plateau in capability pretty soon, and will only marginally improve in the future. Anyone who claims to know one way or the other is lying.

The naysayers out there are quick to point out how easy it is to get tools like ChatGPT to either be wrong, or straight up “lie”, as if it was an actual conscious being already, and therefore can somehow be either “dumb” or “malicious” when it can confidently tell you something incorrect. Many people compare it to the hype of Crypto or Web3, which has never materialized into much more than just money speculation, fraud, and financing criminal activities. I could write a ton about how much of an empty hype Crypto has been, but to me there is a pretty clear and meaningful difference.

I want to start by saying I do actually think AI is over hyped right now. I also think it’s pretty incredible and for me has been extremely useful already, and I’m really excited for its future. These are both true.

Many people know this, but many also struggle with this idea. You have to understand the things current AI is good at, and tailor how you use it to fit that. If you do, I think it can be a very useful tool. Importantly, you have to know and understand that it is not actually knowledgeable, and therefore is not always correct. If you know that, it can still be really powerful.

When I am using AI tools in ways that it is helpful with, it reminds me of Steve Jobs talking about the computer being like a “bicycle for the mind”. It helps me do the things I would normally be doing, but significantly faster and more effective. Here are some ways and specific tools I have found useful -

  1. Summarization - So much of the internet has been filled with absolute SEO garbage, and a lot of times I just need a quick summary of something. Much of the time I have found AI very good at summarizing the key points of something. Sometimes a web page, sometimes a long PDF where I just need some highlights, and many more. It’s important to realize that it’s very possible it will miss something, or be slightly wrong on part of its summary, so if there is an important document like a contract, you probably don’t want to use AI to just summarize it and call it good. But most of the time, I don’t need it to be 100% accurate, and speed can be important. It’s excellent for that.

    I use Readwise Reader as my “read later” app. It has a summarize feature built in that I use all the time. More often than not, the summary tells me if I want to read the entire article or not. It often does actually get me to read a whole article I might not have otherwise.

  2. Coding - I am well aware AI is not a great coder. But I’m not either. I’m in IT, and I dabble in code from time to time. Frankly, I’ve had many developers work for me over the years, and unfortunately most people developers aren’t very good either. But if you know that, it can still be useful. It’s often better than I am, and if you ask things in the right way, it’s honestly pretty impressive. Sure, I’d rather have a highly experienced and talented developer I could just call on any time, but I don’t, so things like ChatGPT and GitHub Copilot have been a great resource for me. Of course I test everything, and don’t just trust that everything it comes back with is correct, and if it’s something high stakes I’d have a person who knows what they are doing review it before making something production, but it has helped me do a lot more than I could or would have otherwise.

  3. Rubber ducking and brainstorming - Somewhat related, I have found it a great resource to rubber duck and brainstorm ideas with. If I code something and I don’t completely understand what made it work, or why something had to be a certain way, I can ask one of these tools, and gain a better understanding. If I’m writing a blog post, particularly for work where we are just trying to establish a presence online, I can bounce ideas off of it really easily. I often find the way it writes to be a little weird or overly formal, but if I take these ideas and run with them, and make them my own, it not only often gives me great ideas or places to start, but dramatically speeds up the time I can do it.

  4. Search - Search is one that it seems like a lot of people struggle with. ChatGPT, for instance, is not really search. There are millions of comparisons to how ChatGPT fails at some search-like thing where Google search does just fine. But that’s not really what ChatGPT is. For this, I use Perplexity AI. Perplexity uses several of these LLM technologies under the hood, but has put them together in a way that honestly has completely changed search for me.

In their own words, this is what Perplexity is -

Perplexity is an alternative to traditional search engines, where you can directly pose your questions and receive concise, accurate answers backed up by a curated set of sources. It has a conversational interface, contextual awareness and personalization to learn your interests and preferences over time.

Perplexity’s mission is to make searching for information online feel like you have a knowledgeable assistant guiding you. It is a powerful productivity and knowledge tool that can help you save time and energy with mundane tasks for a multitude of use cases.

How does Perplexity accomplish this?

With the help of our advanced answer engine, it processes your questions and tasks It then uses predictive text capabilities to generate useful responses, choosing the best one from multiple sources, and summarizes the results in a concise way.

Perplexity to me has been like a research assistant. It is truly AI powered search. It does a fantastic job answering simple questions, but importantly provides links to the references that were used. I’ve seen many online publications panic about tools like this (It’s similar to Arc Search as well, which I use on my phone), because it will often just answer a question, meaning you don’t have to click through to get an answer. While that logic makes sense, I honestly find myself clicking through far more than I ever do with Google.

With Google, so many results are just SEO garbage that increasingly don’t even answer your question. Content farms pump out keyword filled garbage articles, cram as many ads per inch as they possibly can, and game Google’s system very effectively. As I’ve written several times recently, this ruins the quality of results and honestly has made Google search nearly unbearable. Perplexity somehow seems to cut through that junk, and not only answer my question well, but all the links are much much higher quality in my experience. Because of this, I find myself clicking through FAR more regularly to the linked resources for more information.

The downside is that if you need to do more than a couple of searches and you want the highest quality where it searches for you, you have to pay for it, and these tools add up quickly. The time it saves me and how much it helps me with my work make it absolutely worth it to me, but this isn’t going to replace “free” search engines, but this is the first time in many many years where I feel like Google has some legitimate search competition. If you haven’t tried it yet, I encourage you to do so.

I think there will be many more extremely useful applications for AI soon. For instance, once it can more effectively use your own data as a source of information, I think that will be pretty revolutionary.

AI Tools I Use The Most

  • ChatGPT / OpenAI - https://chat.openai.com and OpenAI API for certain scripts, Zapier automations, and more - https://platform.openai.com/

    Note - Using the paid version for GPT 4 makes a big difference in quality of results

  • GitHub Copilot - https://github.com/features/copilot I use my student copy of this with JetBrains to have integration right in my IDE. I have found this slightly less useful than using ChatGPT, but more convenient for specific scenarios.

  • Perplexity AI (Referral link. You get a discount on the first month if you use this link)- https://www.perplexity.ai/