Looking at AI issues popping up recently, and thought it might be interesting to ponder these topics with fellow Wooters. C’mon in.
1 Like
There are going to be some winners and some losers.
Bloomberg
Sam Altman Has a Lot of Things Keeping Him Awake at Night
OpenAI is facing disadvantages on numerous fronts compared with rivals.
https://archive.ph/Bc2uq. (No paywall)
Oops!
Bloomberg
OpenAI’s Latest ‘Breakthrough’ Is a Sobering Reality Check
https://archive.ph/DvAF5. (No paywall)
Your info can be used against you in court. Targeted ads with no ability to opt out.
2 Likes
Smart Homes that automate for their owners.
1 Like
Wow, turn off the electricity!
1 Like
It’s extremely useful and powerful technology. Where I work, there is an internal portal with several models to choose from, and we don’t have to worry about feeding it proprietary information because it’s an internal resource.
I use it all day long at work to help me solve problems that would require hours of research, trial and error. Sure, I could do the same thing myself, but it might take me hours or even a couple of days per problem. Having a machine assistant gives me the freedom to direct my energy and brain power toward the things that matter most. For someone with a work queue that is chronically months out, (the guy writing this) having this kind of assistance is very welcome.
It makes a few mistakes now and then, and I almost always catch them, but the vast majority of the time it’s spot on. The quality of the response you get is proportional to the quality of your prompt. For a complex problem, I’ll write a two paragraph query, feed it technical documents with thousands of pages, and then have a conversation with it about the material until I find or generate the information or data I need.
“Intelligent” isn’t the right word in my opinion, but if we personify it, it’s extremely intelligent. Astonishingly so. Naysayers who claim that it’s all “slop” haven’t accurately seen what it’s capable of.
The Bad News:
It’s not so much the large language models (LLMs) I’m particularly worried about. It’s humanity. Like nuclear proliferation, history has demonstrated that we cannot be trusted with something so powerful.
Now anyone can doctor or fabricate any video and make it look very convincing. We are entering an era where it will be next to impossible to discern “truth” from from complete lies. If you think it’s bad now, just you wait. Fabricated media is a propagandist’s dream, and the quality of the fabricated photos and videos will only improve. Tyrants everywhere will manipulate the masses with ease. Everyone assumes that they are the ones who will never be fooled, but to quote a cliche, a false sense of security is worse than none at all. Get ready for it. You will be successfully deceived at least some of the time.
Cyberbullying (affecting both kids and adults) is going to be soul crushing. People will ruin the reputations of others with fabricated content.
Everyone’s data will be continuously mined as it already is. In the recent past, it took teams of data scientists and some very complex code to predict your next move with accuracy. This is how advertising and Facebook’s algorithms work or worked, but the transition is taking place rapidly.
Billions of dollars change hands to predict that move you’re going to make. It’s a huge and powerful industry. With the AI tech now, it’s much, much easier to analyze millions upon millions of data points and come up with trends.
As a small example, at work I dumped literally millions of database records into a file and fed an LLM the data. I then asked it questions about correlation and trends and it presented me with detailed stats. It would have taken me a few hours to script that myself, and another few hours to mine the data meaningfully with some expensive data analysis software.
You know who else would love to predict your every move? Hostile and oppressive governments.
This is a Pandora’s box, and the decisions that are being made right now with virtually no oversight will affect humanity for generations. Now is the time to urge politicians to take this seriously.
2 Likes
6 Likes
Yeah, I’m sorry, that was really long. Things are happening so fast.
I’m also not holding my breath in thinking politicians are going to do anything useful. I’ve become cynical.
1 Like
Impressive comment, as usual!
How do you know the info is correct?
Poor guy!
They already do!
We can expect skyrocketing suicide rates.
Palantir?
Our own government. Crooks, attorneys, your competition, stock brokers and gamblers would love to to know your next move.
Those dummies don’t understand what needs to be done any more than the common jake. We cannot trust the Billionaires who own the programs to be honest and help with guardrails.
AI use in tech jobs is making it hard for new grads to find jobs.
I wonder what percentage of students just use this AI thing to do all their work?
I see this as the further dumbing down of generations…
1 Like
BINGO. Think of the money $$$ THIS is going to generate. Users should worry about using verified user identity, and the potential of data leaks of those verified user histories. Of course there is the potential of addiction and even more money $$$ generated.
The reader comments were interesting on this article.
This will surely deepen the crisis of loneliness, decreased dating and marriage, alienation, insecurities with intimacy and relationships, etc.
A challenge to Only Fans and the porn industry?
1 Like
Nope, you cannot just turn off the electricity if it escapes onto the open internet.
This is scary!
Here are some practical examples:
Generating Code
If I’m having it write code for me, it’s almost always for the purpose of automating some task that saves me a ton of time. Other times it will be a small part of another application.
My queries about automating tasks often refer to code languages I’m dangerous in, but not proficient. Or, just an awful and cumbersome language like Windows batch scripts.
The accuracy of the response is easy to verify:
Often, I’ll go through the code line by line not only because I’m checking for mistakes, but also because I want to know how to do this myself the next time.
Other times, I’ll simply test the code module 12 different ways to ensure it does what I want it to do. (We call this “unit testing.”) Whether I’m writing a code module or I’m getting AI assistance, I test each module as I go and debug it by itself.
Researching Technical Documentation
In the case of feeding it a huge, thousand page document and a lengthy query, this too is easy to verify. I’ll check its response by going to the document and looking up the sections it referred to. If I’m not sure where it got its answer from, I’ll simply ask it.
If I’m putting the information to use as part of an implementation, I’ll often make a proof of concept to ensure that the information is correct. If it doesn’t seem right, I’ll give it more context about what I’m doing and ask it to clarify. It will then go through the results of the test and conjecture about what might be wrong, or point out which part of my implementation is flawed.
(Nerdy analogy: If you recall Geordi LaForge from Star Trek having problem solving conversations with the ship’s computer, it’s pretty similar to the flow of my conversations with it.)
Once the proof of concept works, I’ll move on to writing production code.
Mining Data
If I’m asking it for statistics from a huge trove of data, this is trickier to verify. That said, when I ask it to go through data like this, I’ll typically be looking for basic trends to point me in the right direction. I won’t use the results for anything mission critical until I’ve verified its conclusions. It’s a major win because it saves me from writing some complex database queries.
1 Like
You obviously don’t work for Deloitte then.
1 Like
Most definitely. If it can fabricate realistic pictures and video, you can be sure people are using it for this.
Alternatively, since an LLM can emulate any kind of conversation or story, it can immerse you in an interactive trashy romance novel or pretty much whatever kind of story scenario a person wants, erotic or otherwise.
1 Like
(Edit: I found an article referencing what you’re probably talking about and now I understand what you’re getting at! I think the latter paragraph below is still very relevant)
No, I do not. But in a scenario of accounting and taxes, for example, AI would excel at finding discrepancies that could then be reviewed by a human.
Besides, if I were writing code for the finance industry, the best practices of software engineering are still the same and I could easily use similar or identical processes without introducing any risk. It’s all about how you use it. If you’re a competent engineer, you’ll use it to competently increase your productivity. If you’re an incompetent or lazy engineer and you use it irresponsibly, then you’ll get yourself in trouble pretty quickly and eventually ruin your career.
1 Like