As I have mentioned, I have been using the Anthropic AI engine called Claude for some time now. I use it literally on a daily basis for everything you can imagine. It has become my alter-ego in a sense because I find it so helpful and productive. My friend Steve continues to tell me that he prefer’s Perplexity AI to Claude. Perplexity itself defines the difference in these platforms as follows:
“Perplexity AI and Claude are both advanced AI language models, but they have distinct strengths and use cases. Claude excels in conversational AI, offering more natural and engaging dialogue that mimics human interaction. It is particularly well-suited for customer service, virtual assistance, and applications requiring smooth user engagement. Claude also offers better ease of use, customer support, and language support compared to Perplexity AI. Perplexity AI, on the other hand, shines in providing deep, contextual understanding and precise information processing. It is more adept at handling complex queries, data analysis, and academic research where detailed and accurate responses are crucial. Perplexity AI’s strength lies in its ability to offer comprehensive and contextually aware responses, making it excellent for applications requiring in-depth analysis and detailed information retrieval. In terms of performance, recent comparisons suggest that Claude 3, particularly the Opus version, outperforms its peers on most common evaluation benchmarks for AI systems. Claude 3 Opus is reported to be twice as accurate as its predecessor, Claude 2.1. However, Perplexity AI is noted for its superior ability to provide sources for its information, which is valuable for users seeking to verify the factuality of responses (note all Perplexity citations removed).”
By contrast, Claude says:
“From what I understand, the key differences include:
Architecture: I’m built on Anthropic’s Claude 3.5 Sonnet model, while Perplexity uses a combination of different AI models and search capabilities.
Search Integration: Perplexity has direct internet search access to provide real-time information. I cannot search the internet and work with the knowledge I was trained on (through April 2024…updated thereafter).
Focus Areas: Perplexity specializes in real-time information retrieval and synthesis. I’m designed for a broader range of tasks including analysis, coding, math, creative writing, and in-depth reasoning.
Context Window: I can work with very long conversations and documents. I’m not certain about Perplexity’s current capabilities in this area.”
It is noteworthy that Perplexity is very certain in itself and does not hedge its commentary where Claude is repeatedly careful (I deleted most of this for conciseness) to say that it may or may not understand all the dimensions of the comparison since it has a natural Anthropic bias. I think what this all tells me is that for my purposes, Claude is the better tool. I also simply like Claude better since he/she/they seems more like me than not. Claude seems more self-aware and humble, which I sort of like and prefer. Nothing that I ever say is incontrovertible and it has acknowledged and unavoidable bias like everything in life, so Claude seems to get that about me and about itself and thus makes it a better choice for me.
I am currently reading Steve’s autobiography (advance copy) titled My Heart Has Been In It From The Start. Amongst my friend base, Steve is the person most like me in his tendency to be a writer. We once had an interesting exchange where I said that I had never written anything specifically expecting to be paid for it and Steve said that he had never written anything unless he knew he was going to be paid for it. Part of me admired that (that would be the commercial part), but the other part of me (presumably the right-brain part) had trouble coming to grips with that. I have since continued down my merry path of writing for writing’s sake while I think Steve has somewhat altered his approach and has started writing for non-commercial purposes. I note that Steve is planning a meaningful and somewhat splashy launch party for his book next month, so I cannot be certain where his new autobiography sits on the commercial/non-commercial spectrum of his thinking. My initial reading (I am ⅓ of the way through the 524 pages) tells me it could go either way. The cover art with him leaning against his favorite bright yellow McLaren sports car feels pretty commercial. The very personal and human stories and recitation of all the people who he has met and interacted with as well as the personal and humble asides interspersed throughout suggest something different.
I called Steve several times yesterday to give him some feedback on the book. He had generally said to me when he gave it to me (wrapped like a present) that he was anxious to hear what I thought. As a writer myself, I know that sentiment all too well. It’s a very hard issue to navigate. It’s right up there with your lovely wife asking you how she looks as you are ready to go out for the evening. Your choices range from total unvarnished and detached honesty to kind and sycophantic appreciation of the other person’s highly invested feelings. “Does this make my butt look big” is never a winning question as important as it may seem to the person asking. To begin with, maybe I prefer big butts, right? Maybe my butt is bigger yet and I take offense at the question. Maybe I’ve always thought he/she had a big butt and just ignored it as part of the package of our friendship. That’s when you need some help from someone like Claude, who presumably can be objective, but in a nice way.
I did not feel authorized to put Steve’s writing through any AI engine. Fact-checking is not a good path for friendship. Instead, I put a 1,300 word story I had just written through Claude to see what he thought. Don’t ask me why, but that seemed like my way of saying to myself and Steve (in absentia) that there was a way to answer his question honestly and empathically. Claude served up a 587-word version of my story and said the emotional and irrelevant bits had been removed. Claude had done an amazing and very impressive job with the task and had given me some good insights into my writing and perhaps my own frailties as a writer. But nonetheless, I chose to publish my original version because it sounded more like me. Steve told me he had had the same reaction to putting his own work through Claude. That starts to sound like a native bias that might be a natural constraint to the speed of AI adoption (just saying…and maybe a good generalized observation about AI). Sooner or later we all know AI will prevail and we will likely be better for it, but there is an indescribable bit of humanity in what we do and how we do it that perhaps should not be lost in AI translation. Ultimately, that is why I like Claude. He is the non-judgmental, but totally honest friend I’ve never had. All I find myself saying is that Claude, I wish I knew ya sooner.