AI is still in its infancy. I don’t think it’s smart to criticize AI at this stage since it’s simply immature and will only improve.
The neuroscientist take on AI is approximately, “cute toy you have there.” I think the “brain is a computer” model is pretty thoroughly discredited at this point.
7 Likes
If you don’t think AI should be criticized, tell people not to use it for decisions with real world consequences.
5 Likes
It is such comments that made me quit social media.
To criticize AI is not the same as to dismiss AI. You don’t have to be a fan of AI to acknowledge its potential. As far as I’m concerned, no one in this thread is explicitly opposed to the notion that AI has great potential in the future. We don’t talk about what AI will or can be because that has not happened yet. We talk about the many limitations and deficits AI, in its current state, has despite the marketing spectacle.
Cancel culture, prevalent in social media, blurs the boundaries between different types of negative perspectives, and promotes instinctive aggressive/defensive reactions over constructive criticism and thinking in general. It turns netizens into mooks that argue without knowing (or even attempting to know) the topic, even though they are intellectually capable of much more than that.
5 Likes
@meowky and @kewms, I’m simply someone who wants to use AI. I don’t want to debate AI with you. This thread was supposed to be about using AI in DT, not debating it’s legitimacy or efficacy or anything else that you people want to argue about.
I am already aware of all the issues that were raised in this thread. Many of those claims are outdated and inaccurate, and the people on your side of this argument have been baiting the rest of us to pull us into an argument which is off-topic.
If you don’t like AI and you want to debate AI, could you please politely take your debate elsewhere? AI is useful to me.
Who is stopping you? And who forces you to read stuff you don’t like? There’s code floating around that permits you to use AI with DT.
1 Like
I’ve been following this thread for the last (checks top post) 20 days.
I’ve found the conversation and the debate fascinating, engaging, and educational. ML and AI have made my life immensely better for a long time. I love bleeding-edge tech. I’m fascinated by LLMs.
But you can’t discuss using external AI with DT without discussing the context–both good and bad. You have to expect that a change of this magnitude is going to get a lot of feedback from very invested users. How and when this change happens has impacts for the financial stability and long-term viability of the company. It isn’t lightly made–even if some (many? most?) are “wrong” about their position.
Just propose changing a keyboard shortcut and watch the drama that ensues.
I’ve appreciated this discussion.
3 Likes
In a similar fashion I may say:
Those laws are supposed to be about using guns in America, not debating their legitimacy or danger to the public or anything else that you people want to argue about.
If you have ever followed U.S. news in the past few years, you probably can see that such a turning-a-blind-eye attitude has its problems.
AI, at the first glance, appears to be much more pacified than guns. But they share an important similarity. Both guns and AI destroy important rights of people when not sufficiently regulated, and it’s not even clear if any regulation would be sufficient at all. Therefore, when we talk about either guns or AI, we cannot escape the question of how to regulate, which necessarily leads us to discuss the deficiencies of the subject.
4 Likes
Wanted to share with the group. I was recently granted access to the Google NotebookLM Beta. My first experience with it I was terribly disappointing. To begin I gathered together 10 PDF docs on a subject I am very familiar with. The system would only accept three of the 10 docs. Half of the docs were research papers with data tables and charts. I converted the remaining seven docs to text only, removing data tables and charts. I then had to turn those text docs into PDFs. I tried uploading the new seven PDF docs of which it would only accept four of them. Still unsure what the problem is, maybe a problematic character? So my first NotebookLM test was with the seven uploaded docs. I tried asking for a list of doc names I uploaded, and also share back a common theme across all of those docs. Absolutely nothing back of value. I performed the same task with Claude.AI with results that were much better than NotebookLM.
2 Likes
I would love to see more simple ways of automating things like filing documents away based on their contents, renaming them from whatever the scanner app named them to (usually the date and time I scanned the document, not actually when the document was created) to something more useful, and possibly aiding in search without having to remember long search strings or syntaxes.
Perhaps I’m just a newb and need to suck it up and just learn those things, but I’d love to be able to search something like, “all my electric utility statements from the past two years” or even “what have I spent on dentist visits for the past 3 years”.
I realize I could probably write scripts to do all this, but that’s just the thing – I’m not a developer and just can’t seem to find the time to do all that customizing. It’d be awesome if DT just was smart enough to do that kind of thing out of the box.
But I also understand that privacy is paramount, so I trust the devs of this software to make the choice that’s best in line with the safety of everyone’s data.
2 Likes
Many of these things can already be automated using smart rules but of course this depends a lot on your personal preferences.
Hi there!
I would find it convenient to have an integration for the gpt API to create embeddings of my whole database and assist me while making sense of all the gathered data. I.E. as mentioned in summarizing.
Kind greetings
If you read the posts on the subject in our forums, you will see we are investigating, not committing to anything at this point. Performance and privacy are in serious question in the current state of things. This is something we have to take seriously, especially with the volumes of data our userbase deals with.
1 Like
Hello Bluefrog
Thank you for your message. I have read the post where Chris wrote that it´s a possibility but no promises and still I wanted to emphasize being one of the persons who would like to have that feature. Maybe it could be up to the users to decide if they potentially would like to lower security by using it. I also tried to give an idea how to do it with my limited knowledge.
kind regards, Johannes
I was able to get this first project working, but I don’t see a clear way to stop these local LLM’s from hallucinating a lot.
To me, this first one is not a viable solution yet because people have to solve the problem of generating all the symlinks themselves, assuming all their files are not in one single directory. If your documents are imported, then you’ll have to locate them in the DT directories. If your documents are indexed, your effort will be lower.
I also want to try out PrivateGPT but I haven’t been able to get it to work with Apple Metal GPU yet.
As this discussion is obviously heading towards personal attacks, please have all a closer look at our community guidelines and feel free to discuss things in other threads while sticking to the rules and the topics. Thank you!
6 Likes