Don't look for a neural network – you already have one on your team
A neural network can pass the Turing test or the handbrake test, but at the same time it will not be able to think, have empathy and make the fastest possible correct decisions. Moreover, the necessary correct decision may turn out to be an exception to the algorithm, and then the AI may cause harm. And okay, if it’s an old invoice or a stupid text about a new phone, but what if it’s a car’s autopilot? We are not yet able to solve these problems (although, fortunately, we are aware of them).
Blindly following technical fashion can one day simply ruin everything that the team has been creating for years. In addition, customizing and adapting a neural network for specific tasks of a company also requires resources, effort, time, and professional developers. Otherwise it will be a very simple toy.
Below I present the tasks and answers of the neural network with my comments. I didn’t go deep into the development jungle and force the neural network to write code, but gave it very simple, trivial tasks that my co-workers and I perform literally on the fly.
Neural network, work for us!
The text itself is bad: tautologies, repetitions of semantic blocks, questions in the text. But that's not so bad. For example, the neural network offers you to familiarize yourself with our portfolio and see successful projects. Formally, it’s cool that she took this into account, but… we don’t have a portfolio section and you can’t really tell us about our projects under NDA. Where did this part come from? Yes, from a bunch of landing pages with similar texts!
Obviously, the text contains the most typical features of CRM systems. Here is a clear example of when an employee’s head would have worked better. I need ideas for new functions for a specific RegionSoft CRM, and even if I list in the prompt all those that already exist, this will not be of much help either, because it would be nice to show customer requirements, analytics on the most used functions, etc. Much simpler , it’s faster and, most importantly, more effective to gather employees for brainstorming and generate a pack of different ideas, from which you can choose 1-2 that you can take to work.
Some names are simply illogical; you can take two at most, and they do not pretend to be any novelty or creativity. I suspect that the idea of the name “March Youth” is rooted in the numerous campaigns of salons “let’s bring back youth by March 8th.”
I won’t say that this is the most clear and simple definition (again, not without grammatical problems). Although in general it’s practically canonical. There are many factual errors in the definition that can be corrected.
Everything is banal: almost any client sees these reasons to use CRM in the first minutes after entering a question about the system into a search engine. They are simply not suitable for the task at hand. But it’s not interesting, and we’ll still clarify the prompt – we’ll ask for non-trivial motives.
One could be ironic about the fact that the security argument is non-trivial for most CRMs, the neural network clearly knows something 😉 And so – again, a rather simple and well-worn set.
The answer from technical support is formally not bad, but again, it’s vague and is more likely to irritate than calm. Alas, verbiage will not help the matter. But in essence, of course, nothing concrete: offering an overview instead of pointing to the point of the problem and a quick solution is definitely not the right option. No solution is proposed, but at the end the AI clarifies whether it is suitable. However, from experience, more than half of technical support responds something like this 🙂
There are almost no comments here: in some places, apparently, our own advertisements were simply parsed. There are questions about small things, but perhaps this is the most successful task (however, some Direct ads are not an example, they need to be corrected).
This task failed, but with a twinkle: you can laugh heartily. I just can’t imagine what will happen if managers actually adopt the proposed options. Although you also hear this on the phone – so what, bots don’t blush.
Everything is simple here: the neural network does not understand what a commercial proposal is and how it is compiled. Well, the text itself simply does not stand up to criticism in terms of the level and number of cliches.
Please write it shorter, because the text in a commercial proposal is not the main thing. Well, this is already a complete failure. Even the most backward companies do not use this.
We make a clarification and give it another chance – another text about nothing. That is, we need to explain to the neural network what a commercial offer is – and it’s unlikely to cope if we only write a special application for generating texts for commercial proposals.
conclusions
The conclusions that are emerging so far.
Working with prompts for generative neural networks is a separate complex task that takes a lot of time and requires taking into account many factors. It seems that coming up with your head is faster.
There are many factual flaws in the generated texts that sound “on line”, but at the same time spoil the impression of any professional.
The commands are taken too literally (which is expected).
The texts contain a huge number of templates and cliches that you don’t want to see in any corporate content.
Now it’s difficult to say whether the future belongs to neural networks or not. Rather, in this direction, the future still belongs to people who design and write algorithms, take exceptions into account, train neural networks, and fight artifacts and hallucinations. It’s too early to use neural networks in production – and I think that users will quickly get tired of such texts, and the value of experienced leather copywriters will increase. After all, speech is a process that is difficult to fit into algorithms. Of course, if it is competent, beautiful speech from an educated person.
