Welcome to today’s daily kōrero!
Anyone can make the thread, first in first served. If you are here on a day and there’s no daily thread, feel free to create it!
Anyway, it’s just a chance to talk about your day, what you have planned, what you have done, etc.
So, how’s it going?
Thought for the day
What role do you see AI playing over the next 5 years?
Now that the crazy hype train has mostly run out of steam, what do you think that the realistic trajectory of AI integration into our daily lives will mean.
I’m an optimist, I see AI as a great tool to accelerate the power of work and education; but there are massive dangers to make even more addictive “social” media, creating the perfect “opiate for the masses” type of situation. I’m really happy to see the growth of open LLM’s rather than just the closed source ones.
I see huge potential in large-scale data analysis, particularly in disease research.
deleted by creator
I think it will be very disruptive - both in good and bad ways. Part of my new job is infact investigating and implementing ML systems to bring in AIOps to our IT systems and teams - such as using tools which can monitor our infrastructure and automatically resolve issues, or even prevent them from occurring in the first place. Hest yesterday I saw Nokia demo their new generative AI assistant for network devices (which is a part of their new Linux network OS) and that was really cool. So it’s pretty exciting being in the most of all this.
Among the general positives, I think AR/multimodal AI would be an incredible quality-of-life enhancer for some folks. For instance, imagine a blind person wearing an AI powered glasses that can analyze and audibly describe the environment to the user in real-time. Now something like this can already be done today, but it would rely heavily on the cloud and/or not be real-time, but with growth of AI-accelerator chips, it’s not far away where this sort of real-time processing can all be done on the “edge”, in a small form factor like a pair of glasses.
Another application could be in the medical and fitness industries, already we’re seeing micro implementations of AI in various places, for instance smartwatches are already using statistical analysis based on your body parameters and current health metrics, like it can recommend how much you need to walk in a day, how much sleep you need etc all fine tuned for you. But in the future, with more sensors, more processing power and advanced ML models, I can imagine these sort of features can become a lot more powerful - maybe we could see some early warning detection mechanisms for various medical conditions. You could get a complete exercise routine from morning till evening, tuned specifically for you, adjusted in real-time according to your heart rate and pace etc. Or imagine a health provider running an AI which could analyze your entire medical history, maybe combined with your smartwatch data, to paint a comprehensive picture of your health and catch things that your GP might miss.
So when you consider stuff like that, I think it’s very exciting. There will be a few pain points of course such a job losses, privacy issues and lots of legal challenges too, but these issues aren’t insurmountable and we will get over them.
I think education and training will be where we see the biggest impact, as long as we can implement it correctly.
Imagine an expert teacher, specifically tailored to your child, who knows exactly what they have done in the past and how they are tracking. Schools can focus on group work and the emotional needs of kids that allow collaboration and effective working together.
When you are doing individual work; your AI tutor would be able to specifically meet your needs.
Healthcare / diagnostics has huge potential - “AI” is really just advanced algorithms at this point and that’s exactly what drives a big part of healthcare. (Further) Automation of screening tests, combined with AI matching up patient histories etc could change the face of medical treatment, cut waiting times, improve outcomes, speed up diagnoses.
I’m cautiously optimistic!
I am also mild optimistic about its use in aged care, there are a lot of lonely old people in rest homes. I know it is sad and not ideal, but we don’t have the resources to mentally stimulate them all.
I see an advanced AI chatbot with a voice plugin module, used with wireless headphones. There could be a much better effort to mentally engage with the older folk; I know before my grandad died, he was stuck in an aged care facility, I didn’t get to see him that often, only when I was in the same city as him. He had a lot of very lucid days, even right up to the end, but he was bored.
It would be quite nice to know that people even if they are not physically able, can still mentally engage. The chatbot could talk to them whenever they felt like talking, it could monitor symptoms etc…
I know that old people currently use the doctor here as an excuse to have a chat, and that is telling.
I’m not sure we will see a lot in the next 5 years. ChatGPT is not at it’s level because of advances in this area, but because they took the dive and consumed huge amounts of copyrighted material to train a model described by Google in about 2017. With the internet now flooded with AI generated content, the data can only get worse so any significant gains need to come from the model.
At some point we will see rapid gains once machine learning gets good enough to write a model better than what we have, and then uses that better one to make an even better one, and repeat. But I don’t think that’s in the next 5 years, but there’s a good chance it’s within my life.
The hosting itself is the major cost and this is provided by fediservices.nz, you can donate to them here.
It’s not hosted on a commercial VPS but on a machine in a datacenter in Auckland that also hosts other services they provide, so the costs are pretty fixed and mostly relate to hardware (as an example, they’ve recently added more storage specifically to address storage requirements for Lemmy, which is growing a lot because the image cache is not configurable, and cached thumbnail images are full resolution not resized. The next release should help resolve this by allowing configuration of how long the cache is kept or if it is kept at all). Currently the image cache is around 150GB.
I pay the domain and email sending costs out of pocket but it’s pretty minimal.
Back in the office today after a busy weekend away at a Cave rescue course in Takaka. Loads to do, but the sun is shining and I have a Real Fruit Ice Cream machine about two paces from my office door :)
