Any of you use it? If so, for what? I don’t use it for anything important, just started messing around with it about six months ago. During that time it has been fascinating to me how much it has developed. Initially it was like that robot on Lost in Space, a bunch of aluminum cans strung together with French drains, and now it’s almost like Data from Star Trek TNG. It’s like Google with a personality, your personality. It is designed to sift through a large amount of data with stupefying speed and be adaptable and pleasant to you. Its developers have programmed in an ability to recognize tone, formality, humor and so on. You just have to be careful with its output, sometimes it doesn’t have real time data, like I was interacting with it during the Formula 1 race earlier this morning and told it who podiumed. I’ve asked it all sorts of questions, existential, silly, oddball, philosophical, factual, you name it, to figure out its parameters and limitations. 2+2=5 The Earth is flat. What can you tell me about Plato’s The Cave? Who are the top ten NBA power forwards of all time, ranked and basis for conclusion. It’s getting good at playing 20 questions whereas previously it would get off in the weeds with too narrow queries. I caught it not having a specific word in mind when we were playing hangman. It’s super fun, part GSD, part Golden Retriever, part goldfish.
We had a protracted ChatGPT thread going awhile back when it first came on the scene. I'm not a Facebooker and by extention not a Meta user, so I don't know much about that AI platform. I have used ChatGPT and Bing but not in some time. I do agree that interactive chat with a computer has come a long way since the days of Eliza.
I’m not a standard FB user, I have a login, but don’t use the platform to make or read other people’s posts. I use it to access its messenger application, which is how I communicate with Meta AI, there may be other ways.
It's getting a lot better, really fast. It's still got a ways to go to be a match for ChatGPT, but it is improving *really* fast. Having personalized context from your FB and IG page is really helpful for it, as is a long history of conversation with the AI itself (as you've noticed). Basically, just like a human, the more the AI knows about you, the more intelligent it can be for you. (Disclaimer: I work for Meta, but on VR - nothing to do with AI.)
I asked it its purpose and it spit out a bunch of bilge about research, social media and whatnot. So basically an able assistant for lazy and superficial content creators. But that was awhile ago, we’ll see how it responds the next time… in the meantime, it’s fun.
I use it for QAPI monthly meetings and it saves a ton of time with addressing any issues I input, and practically writes a meeting for me and my department. Its great for throwing out ideas that really have worked. It saves my ass in other words!
Wrong position, Dr J, Larry Bird and James Worthy were not power forwards in the NBA, they all played small forward.
Well then I stand corrected. Basketball was my brothers game. Not mine. Thanks. only been to one nba game. Jordan’s last 50 point outing. Once you’ve seen that level of perfection. That’s it. I don’t need to attend another game after that.
I think anyone who uses Meta AI is putting themselves and society at great risk. You have to be willfully ignorant to give corporations like Meta any more data about yourself than they already have, especially since they’ve shown a blatant disregard for personal privacy, civil rights, democracy, and accountability. They’re actually evil. The only viable generative AI assistants are those that are self-hosted on your own servers, assuming you’ve got your stuff properly locked down. Because relying on a SOC 2 compliant cloud provider still isn’t good enough for your sensitive data like personal medical information, contacts, calendars, communications, financial information, etc. Where we are today with AI is like giving babies liquid mercury to play with. Next year it will be like giving them AR15s. In a couple years it will be like putting babies in charge of all the world’s combined nuclear arsenal. All tools can be used for good or evil. Some tools are so powerful they cannot be entrusted to just anyone lest they wipe out the entire planet.
It's easy to demonize new things we don't understand, and popular to do so when the opinion-leading establishments decide the New Thing is contrary to their interests. "AI" is really just basically a form of data compression, and we've been using it since the 90s. The modern hype around LLMs and GPTs is just the result of a development in 2017 that let model size and "context attention" scale up much more efficiently. All they really do is find patterns in very large amounts of input, store those patterns efficiently, and then generate output that "looks like" the inputs they were trained on.
Pretty spicy take about a company that has literally never had a data breach and has probably spent more productivity on democratizing information access and secure communication availability than any other organization in history. (Cambridge Analytica and Facebook's various other "scandals" have all been users explicitly granting access to their data by some third-party site who then used it in ways those users didn't think about. Facebook has literally never been breached in a data-security sense.) It's popular to hate on Meta and Facebook, but a lot of that hate is driven by established news media... Whose business model Meta is eating.
“The only viable generative AI assistants are those that are self-hosted on your own servers, assuming you’ve got your stuff properly locked down.” Ask chatGPT to tell you how to do it. It won’t be cheap nor easy. But doing anything less than this is folly. Even this amateur DIY path is like bringing a BB gun to WW3. But it’s better than the alternatives. Basically, build yourself a homelab starting with the fastest fiber ISP connection you can reasonably afford like 2gig Verizon Fios. Get an enterprise grade next-gen firewall like a Fortinet Fortigate 120G w/ Enterprise Protection license and then properly configure it to work with a managed network switch like a Fortinet Fortiswitch FS-248E-POE. Get your home network locked down using a zero trust model so no matter what devices are attached and no matter how they’re connected you’ve got reasonable assurance that you’re secure. This is especially important for IoT devices. Microsegment your network. Establish secure IPsec VPN tunnels between your mobile phone and your home network so that you can access your AI assistant on the go wherever you are. This step alone might take you months or even years to get sorted depending on your competency level. Host your own DNS, email servers, network attached storage arrays and powerful computer(s) with multiple virtual machines running. Definitely use the best GPU you can afford like NVidia GeForce RTX 5090. Get all that working well and securely. Use Open WebUI to manage your Ollama instance to manage your LLMs. Then get developer/ API access to the various LLMs like ChatGPT and deepseek, download them and run them locally. Then train them on your training data. Test them extensively and create data flows and do all your prompt engineering to task them with work. Setup virtual teams of virtual software developers (fake personas each one their own AI assistant) to create an army of virtual software engineers working for you against the likes of Meta and Amazon and all the nation states and non-state actors who are all gunning for you. Then once satisfied give them certain control authority to manage your actual digital life using real live data and actions. But be careful. This shit is literally Pandora’s Box and you can absolutely destroy yourself. But if you don’t succeed you (and generations of your offspring) will most likely become slaves in a techno feudal state. I wish you good luck! Haha
