Please define what a “token” is without AI jargon. And I also hate the word “utilize.” That’s about all I agree with in your article.
Unfortunately your AI evangelism is a reflection of but one more tech solution in search of a problem. XXI C. tech advance lacks the organic quality of real people in real organizations needing solutions to real
problems they have encountered in real life. You and the tech bros seem to believe in a number of things that have never really been agreed upon by everyone else: that new engineered capabilities in the arena of processing more data, faster and more cheaply, always and everywhere produces more social and economic welfare and positive outcomes. Sort of like an ultimate free lunch. I’m a PhD economist and that is the most debatable of all your hidden propositions.
In the case of your writing process, and this article, I am certain that without a GenAI bot, you could have produced just as effective a display of your fundamental unexamined biases and uncritically held assumptions as you did with the help of one. And so it will be with automated help, if you fail to abandon AI evangelism in favor of rigorous scientific disinterest and dispassion. Such would require you to examine all your biases in the context of the social and economic welfare concerns I have alluded to, with emphasis on the distinction between efficiency—more and faster—and effectiveness, a human quality.
In full disclosure, I did not use a GenAI bot to read your article, think about it in the context of the other AI articles I have voraciously devoured in recent weeks, and compose my critique of the output of your AI-assisted writing.
Thanks for your feedback! I read it (without the help of AI) and I really appreciate it. I respect every take, opinion, and criticism.
As for your question, the reason we use the word "token" instead of something else like "word" is that a token can either be one word, part of a word, or any other kind of data (molecule composition, pixels, sound, etc.).
Actually, many people don't want to learn how to use prompt words; they just input something casually and then, if they are not satisfied with the results, they stop using them.
Learning any tool involves a process and methods, which is why my first column in the pub explains how to write prompt words. If you haven't even mastered the most basic methods of use, then your daily work methods will remain unchanged. You are doing repetitive work.
I totally agree. Part of it is due to the hype in the media that portray these technologies that can do anything. Lay people come to them expecting them to magically do anything there is. But the truth is, to get serious things done, you need to do careful prompting. I'm always working on improving and updating my prompt templates
I think - as a user rather than an 'observer' - of these tools that experimentation is very important. And it becomes even more so for those 'consumers' who do not often or who have never been exposed to these tools. Especially in organizations, we stop at uses that are now redundant and reiterated, but we could understand how to adapt specific bots or applications to the tasks and analyzes to be done. For example, I'm also experimenting with tasks like finding things to do in NYC and so on. Thanks for reiterating that.
Does your process include follow up conversations or are you tuning your prompt to get the desired response all in one go? I find giving an LLM feedback to their response often works well - sometimes it goes many turns.
I tried that format. Ultimately, the best thing that worked for me was crafting a prompt that gave me a decent draft that I could then improve through further editing.
Please define what a “token” is without AI jargon. And I also hate the word “utilize.” That’s about all I agree with in your article.
Unfortunately your AI evangelism is a reflection of but one more tech solution in search of a problem. XXI C. tech advance lacks the organic quality of real people in real organizations needing solutions to real
problems they have encountered in real life. You and the tech bros seem to believe in a number of things that have never really been agreed upon by everyone else: that new engineered capabilities in the arena of processing more data, faster and more cheaply, always and everywhere produces more social and economic welfare and positive outcomes. Sort of like an ultimate free lunch. I’m a PhD economist and that is the most debatable of all your hidden propositions.
In the case of your writing process, and this article, I am certain that without a GenAI bot, you could have produced just as effective a display of your fundamental unexamined biases and uncritically held assumptions as you did with the help of one. And so it will be with automated help, if you fail to abandon AI evangelism in favor of rigorous scientific disinterest and dispassion. Such would require you to examine all your biases in the context of the social and economic welfare concerns I have alluded to, with emphasis on the distinction between efficiency—more and faster—and effectiveness, a human quality.
In full disclosure, I did not use a GenAI bot to read your article, think about it in the context of the other AI articles I have voraciously devoured in recent weeks, and compose my critique of the output of your AI-assisted writing.
Thanks for your feedback! I read it (without the help of AI) and I really appreciate it. I respect every take, opinion, and criticism.
As for your question, the reason we use the word "token" instead of something else like "word" is that a token can either be one word, part of a word, or any other kind of data (molecule composition, pixels, sound, etc.).
Actually, many people don't want to learn how to use prompt words; they just input something casually and then, if they are not satisfied with the results, they stop using them.
Learning any tool involves a process and methods, which is why my first column in the pub explains how to write prompt words. If you haven't even mastered the most basic methods of use, then your daily work methods will remain unchanged. You are doing repetitive work.
I totally agree. Part of it is due to the hype in the media that portray these technologies that can do anything. Lay people come to them expecting them to magically do anything there is. But the truth is, to get serious things done, you need to do careful prompting. I'm always working on improving and updating my prompt templates
I think - as a user rather than an 'observer' - of these tools that experimentation is very important. And it becomes even more so for those 'consumers' who do not often or who have never been exposed to these tools. Especially in organizations, we stop at uses that are now redundant and reiterated, but we could understand how to adapt specific bots or applications to the tasks and analyzes to be done. For example, I'm also experimenting with tasks like finding things to do in NYC and so on. Thanks for reiterating that.
Does your process include follow up conversations or are you tuning your prompt to get the desired response all in one go? I find giving an LLM feedback to their response often works well - sometimes it goes many turns.
I tried that format. Ultimately, the best thing that worked for me was crafting a prompt that gave me a decent draft that I could then improve through further editing.
Follow ups work really well for coding tasks.