Today, I wanted to post a short line on Twitter. Someone had asked which was the last book other users had read about a particular topic, and as I casually happen to be reading a book connected to that topic right now, I was keen to answer—I usually don't fall for such things, since I know people with verified accounts keep dropping saucy or gripping questions and making provocative remarks simply as baits to get more interactions. However, chatting about books and movies and making lists thereof is a weakness of mine—so every now and then I swallow the bait, so to speak.
Hence, I decided to post the cover of the book, adding in French (the original question was in this language) something like "I'm reading this right now". My guess was "En le lisant, maintenant", but as my spoken French is a rather faltering one, I decided to ask DeepSeek. The answer was incredibly swift, as usual (I confess I still don't seem to get used to this AI immediacy). He/she (?) said that my phrasing wasn't technically wrong, but that it didn't sound like fluent, natural French, and proposed "En train de le lire" or "Je le lis en ce moment" instead.
I have to say it wasn't my first interaction with HAL 9000—I mean, DeepSeek and/or ChatGPT—but the device keeps impressing me any time I try it. I cannot help wondering about the social effects of this tool's existence. Perhaps I should ask him/her about this.
However, I keep withstanding the impulse to ask for corrections of longer texts. That's what Journaly is all about, isn't it? I still prefer human interaction, mind you. (Wait, no bots around, huh?)
I haven't tried DeepSeek, but since ChatGPT is often times busy when I want to ask something urgent, I might want to give DeepSeek a try. By the way, I love the header image!🥖
Glad you mention it. I made it with Tengr.ai, an AI image generator :P
I didn't find many differences with ChatGPT. However, I prefer using DeepSeek lately because of its being open-source, as they say
@eugen_blick : Lately, I have been using Artificial Intelligence technology to work on musical scores. The result is a bit discouraging. Neither chatGPT, nor DeepSeek manage well the scores in pdf format nor their export to MusicXML format, among others. So we will have to wait and, for now, make do with general ideas for interpreting musical scores.
On the other hand, I am trying to create a semantic search engine with two operating systems on Ubuntu, a Linux distribution. The result is in favour of DeepSeek because it is much more precise and accurate.
Finally, for writing, translating and tasks related to literary activities, artificial intelligence is only useful to take away my fear of the ‘empty page’.
Hi @druida. Could it be there's some platform more apt to perform musical tasks? Whatever the case, it will eventually show up, or ChatGPT /DeepSeek are likely to develop such abilities in time. I find simply amazing the reports I find on the net, many of them speaking highly of the way some or other platform performs very specific and difficult tasks. Take this for instance https://x.com/DeryaTR_/status/1886958462321471708
Hola, @eugen_blick , de nuevo y gracias por compartir. Es cierto que la Inteligencia Artificial va a transformar como se dice la forma en la que trabajamos y pensamos pero, desde mi punto de vista, todavía, estamos en los prolegómenos de esta tecnología. Si el prompt/enunciado es muy generalista, la IA nos da una respuesta más o menos aceptable pero si es muy especializada, la cosa cambia y habrá que esperar. En los campos en los que yo he probado (música, imagen, programación, etc) los resultados no son tan espectaculares aunque sean sorprendentes...
Al ser una tecnología nueva, estoy investigándola con mucha curiosidad y para mí el “bocato di cardinale” son los "embeddings", la clave de este andamiaje...
Te mando un enlace por si estás interesado: https://openwebinars.net/blog/embeddings/
Interesante, después profundizo. Gracias!
This is an interesting thought - I think one of AI's limitations as a language tool is that it relies on machine learning. It can only imitate what already exists, and so it cripples our ability to be creative/innovative with words. I'm worried that the more we rely on AI, the more our range of thought and expression will narrow. That being said, it's excellent for translation and practical use, and it's saved me from awkward Spanish phrasing many times.
Eugen, as far as I know, DeepSeek has only been trained only on English and Chinese, so you can draw your own conclusions about Deepseek's French.
@Lariza Try out claude.ai. The 3.5 Sonnet model is just as good (or as bad) as ChatGPT's top model. If you run out of daily free usage, try also duck.ai. It allows you to use five different models for free:
Of course, duck.ai has also a daily usage limit, but it's really loose. You can ask dozens and dozens of questions before you reach your daily limit.
Yeah, Simone. I noticed there's a good deal of hype about AI these days. I take everything I read with a grain of salt
Oh, thanks so much, I'll take a look! @Simone-