[TechSpot] Artificial plant purifies indoor air and doubles as a power source for small electronics (beats real plants and traditional purifiers)

[Shihab]

Reducing artificial intelligence or machine learning to LLMs is possibly the worst thing to come out of the on-going farce (edit: referring to the chatbot craze in the tech/internet industries), imo.
No. A machine that can talk back to you or that which can read and summarize a few thousand texts won't do much more than that, for similar reasons to why poets and fiction writers aren't crowding CERN, AGU conferences, or whatever. Humans aren't languages. Language is only part of what we are.

 

[Space Lynx]

Reducing artificial intelligence or machine learning to LLMs is possibly the worst thing to come out of the on-going farce (edit: referring to the chatbot craze in the tech/internet industries), imo.
No. A machine that can talk back to you or that which can read and summarize a few thousand texts won't do much more than that, for similar reasons to why poets and fiction writers aren't crowding CERN, AGU conferences, or whatever. Humans aren't languages. Language is only part of what we are.

that's simply not true though, a LLM can compute all of that knowledge way better than a single human can, AI has already been used to improve concrete for example, which humans alone were unable to do, simply because the amount of variables you can combine on the periodic table/amounts, etc

 

[TumbleGeorge]

When it comes to making money, and in the west it's all business, trust no one. Including influential figures and even scientists.

 

[lilhasselhoffer]

Reducing artificial intelligence or machine learning to LLMs is possibly the worst thing to come out of the on-going farce (edit: referring to the chatbot craze in the tech/internet industries), imo.
No. A machine that can talk back to you or that which can read and summarize a few thousand texts won't do much more than that, for similar reasons to why poets and fiction writers aren't crowding CERN, AGU conferences, or whatever. Humans aren't languages. Language is only part of what we are.

that's simply not true though, a LLM can compute all of that knowledge way better than a single human can, AI has already been used to improve concrete for example, which humans alone were unable to do, simply because the amount of variables you can combine on the periodic table/amounts, etc

Do you read the articles you link and refer to? I ask because the concrete thing is not what you think it is.

Urbana concrete

The goal set forward was to take a large data set, with compressive strength data. You then iterate variations of the combination of those data until you have a concrete that can have the same compressive strength with a lower carbon emission level. It wasn't AI...it was iterative experimentation with a large data set...technically something that people have done for years (why else do you think we have hundreds of steel grades?).


It wasn't an AI using an LLM...it was a calculator iterating until a desired value was found. So we are clear, that same "AI" is literally useless in all other instances because it has one calculation. All A, no I. Of course this thread is about cyanobacteria that can generate power that couldn't power an LED, that in a closed environment can convert CO2 into energy, oxygen, and water. It's almost like instead of considering how much "it's cool science and the press are idiots" you just post stuff.
Side note, I live in North Carolina. You sound a lot like Woody, a local DJ...Woody and Wilcox show. People actively hate him because he pulls the same "ain't science cool" media articles...and after a certain amount of time you get numb to someone who is enthusiastic but doesn't filter for what is genuinely cool.

 

[Space Lynx]

Do you read the articles you link and refer to? I ask because the concrete thing is not what you think it is.

most of the articles I read yeah, sometimes I may get it wrong though, just like actual scientists do from time to time

 

[Shihab]

that's simply not true though, a LLM can compute all of that knowledge way better than a single human can, AI has already been used to improve concrete for example, which humans alone were unable to do, simply because the amount of variables you can combine on the periodic table/amounts, etc

Ability to regurgitate words does not make one a innovator.
Now, when we do get to the mythical AGI, distinction between NLPs and the rest may be moot (but even then, I'd wager, a specialized tool would be more effective and efficient). But right now, they aren't.
There are many machine learning algorithms used in science and engineering. Some are used in actual production, not just hypothetical studies. Land cover/use classification using satellite imagery has been, for a while now, typically carried out using some ML algorithm; Random forest, SVM, k-means, whatever. They may share some similarities to your popular LLMs, but they are not the same.
Perhaps someone could train an LLM to translate the output of LCLU change analysis for use in some decision support system, but unless things have changed while I wasn't looking, it wouldn't be used to do the classification itself.

 

[Endymio]

(image deleted)
When it comes to making money, and in the west it's all business, trust no one. Including influential figures and even scientists.

I'm unsure of the point you're attempted. Doramad wasn't marketed or promoted by "influential scientists" -- it was snake oil, no different than the vitamin and supplement industry of today. The uneducated bought it because they falsely believed that trivial amounts of radiation were good for health, just as those same uneducated buffoons today believe that trivial amounts of radiation are bad for their health. The pendulum swings back and forth, but never comes to rest on sanity.

Fun fact: the vast majority of so-called "health springs" that people have for centuries bathed in -- and bottled drinking water from -- have elevated levels of radioactivity, sometimes highly so.

 

[JWNoctis]

Ability to regurgitate words does not make one a innovator.
Now, when we do get to the mythical AGI, distinction between NLPs and the rest may be moot (but even then, I'd wager, a specialized tool would be more effective and efficient). But right now, they aren't.
There are many machine learning algorithms used in science and engineering. Some are used in actual production, not just hypothetical studies. Land cover/use classification using satellite imagery has been, for a while now, typically carried out using some ML algorithm; Random forest, SVM, k-means, whatever. They may share some similarities to your popular LLMs, but they are not the same.
Perhaps someone could train an LLM to translate the output of LCLU change analysis for use in some decision support system, but unless things have changed while I wasn't looking, it wouldn't be used to do the classification itself.

I think you might as well be describing 90%+ of current scientific output, too.

LLMs do have one thing going for them: They have become good enough to sound comptently human to the reasonable person. This kicked anthropomorphic bias to high gear and led to people thinking they would replicate and replace human labour; Whether they could actually do that efficiently and without competent supervision - e.g. code without a competent programmer doing the context and architecture, and verifying the output - remains to be seen.

They are also displacing low-level labour in software industry, and related image-capable models are doing much more of the same in art industry. Both are causing actual, human, harm right now and most possible outcomes going forward are horrifying.

They also already worked admirably well as a Lotus-Eater Machine with the right prompt. That was the magic behind early killer apps like AI Dungeon 2/3, though it seemed to have mostly faded into the background these days, somehow. There might still be room for some popular breakthrough there.

I'm unsure of the point you're attempted. Doramad wasn't marketed or promoted by "influential scientists" -- it was snake oil, no different than the vitamin and supplement industry of today. The uneducated bought it because they falsely believed that trivial amounts of radiation were good for health, just as those same uneducated buffoons today believe that trivial amounts of radiation are bad for their health. The pendulum swings back and forth, but never comes to rest on sanity.

Fun fact: the vast majority of so-called "health springs" that people have for centuries bathed in -- and bottled drinking water from -- have elevated levels of radioactivity, sometimes highly so.

To be fair, history is replete with examples of experts famed and skilled in their own fields promoting things beyond their own expertise for other reasons. People doing thinking for a living are still people, and are subject to the same foibles, too.

At least (almost) nobody advertises high radon content in those spring water anymore.

 

[Endymio]

They are also displacing low-level labour in software industry, and related image-capable models are doing much more of the same in art industry. Both are causing actual, human, harm right now

And electric lights are causing actual human harm to whalers who have no one to sell their oil too, and mechanical looms are causing actual human harm to weavers. I hear they're even working on a contraption that will allow people to ride a machine like a horse ... what do you think that'll do to all the grooms, farriers, blacksmiths, trainers, and horse breeders in the country? Horrifying thought!

At least (almost) nobody advertises high radon content in those spring water anymore.

But the radon's still there, and people still bathe in it. As do people who move from a city like Miami to one like Denver, increasing their radioactive dose by several hundred millirem each and every year ... then freak out at a chance exposure to a few picorem.

 

[Count von Schwalbe]

I think you might as well be describing 90%+ of current scientific output, too

The reason being because in terms of volume that's probably the amount of "current scientific output" generated by LLMs.

or if nothing else, maybe it will give some other scientists some ideas on how to branch out innovative/creative filtration systems

Biochemistry is inherently messy, only good for more biochemistry. Unless they manage to figure out an artificial photosynthesis without biological input, this isn't going very far.

Anyway, if you were wondering why the power output was so low, it was due to the high amount of energy sequestered as carbohydrates. If it output as much energy as it captured on photosynthesis it wouldn't reduce CO2 levels.

I wonder if the apparently high CO2 sequestration of cyanobacteria could be used for atmospheric reduction, or if the concentration would be too low. Could be an alternative source of biofuels vs using traditional crops.

 

[yfn_ratchet]

The device can remove 90 percent of indoor carbon dioxide, reducing CO2 levels from 5,000 ppm to 500 ppm. In comparison, biological plants achieve only a 10 percent CO2 reduction. This technology can significantly improve air quality while also producing enough bioelectricity to power portable electronics.

Random numbers like that tend to raise my doubt hackles. Similarly, it's a controlled, indoor environment, far removed from the entropic outdoor environments natural oxygen factories would be adapted to, especially conditions like a rainforest or algal bloom. If this kind of plant can be leveraged in outdoor conditions to augment current carbon capture tech, color me intrigued then. As of now, this sounds like the latest edition of '2 undergrad students at MIT discover revolutionary battery chemistry' whose results are never able to be replicated.

 

[JWNoctis]

And electric lights are causing actual human harm to whalers who have no one to sell their oil too, and mechanical looms are causing actual human harm to weavers. I hear they're even working on a contraption that will allow people to ride a machine like a horse ... what do you think that'll do to all the grooms, farriers, blacksmiths, trainers, and horse breeders in the country? Horrifying thought!

All fun and dandy until one of these things happen: AI/ML largely remains at current level and start eroding human skills since it took the easy jobs and much fewer could get into the industry, or AI/ML replaces human in an economy with far less capacity to absorb manpower with redundant skills than back during those times you noted, or things happen such that it is no longer economical or feasible to run these models with all the associated suffering of whatever that would also imply.

But the radon's still there, and people still bathe in it. As do people who move from a city like Miami to one like Denver, increasing their radioactive dose by several hundred millirem each and every year ... then freak out at a chance exposure to a few picorem.

People being people and having disproportionate reactions to things they perceive over things they don't. Yep.

 

[Endymio]

Random numbers like that tend to raise my doubt hackles.

Yes, that claim was so obviously sketchy that I initially thought some journalist had garbled it along the reporting food chain. But it exists in the research paper itself. A few house plants if left alone in a building will eventually reduce a 5,000 ppm level down to about 200 ppm, at which point photosynthesis stops -- so obviously a time factor must be involved.

In reading the paper, I see their "90% reduction" was primarily achieved through surface adsorption, not photosynthesis -- which means it's a one-time effect. How well this will stack up over the long term compared to natural plants remains to be seen.

All fun and dandy until one of these things happen: AI/ML largely remains at current level and start eroding human skills since it took the easy jobs and much fewer could get into the industry, or AI/ML replaces human in an economy with far less capacity to absorb manpower with redundant skills than back during those times you noted, or things happen such that it is no longer economical or feasible to run these models with all the associated suffering of whatever that would also imply.

An impressive word salad. Devices like the steam engine and electric motors and water pumps have eroded man's ability to dig ditches and plow fields by hand, as well as manually carrying water from the town well back to the hut. Why, housewives of yesteryear used to be strong as oxes, and with good reason. Should we return to those days to avoid losing essential skills to technology?

 

[Space Lynx]

perhaps this isn't what I thought it was then, none the less I think if we as a species are to ever break boundaries, it requires creative ideas like this that lead to other creative ideas, and so on and so forth

 

[JWNoctis]

Yes, that claim was so obviously sketchy that I initially thought some journalist had garbled it along the reporting food chain. But it exists in the research paper itself. A few house plants if left alone in a building will eventually reduce a 5,000 ppm level down to about 200 ppm, at which point photosynthesis stops -- so obviously a time factor must be involved.

In reading the paper, I see their "90% reduction" was primarily achieved through surface adsorption, not photosynthesis -- which means it's a one-time effect. How well this will stack up over the long term compared to natural plants remains to be seen.

It actually was adsorption plus photosynthesis, though the volume was small.

An impressive word salad. Devices like the steam engine and electric motors and water pumps have eroded man's ability to dig ditches and plow fields by hand, as well as manually carrying water from the town well back to the hut. Why, housewives of yesteryear used to be strong as oxes, and with good reason. Should we return to those days to avoid losing essential skills to technology?

Sorry about that. Guess it came out wrong.

But by that logic if you see humanity itself in general being displaced by something superior as an acceptable outcome, then...you are entitled to your opinion.

 

[lilhasselhoffer]

Yes, that claim was so obviously sketchy that I initially thought some journalist had garbled it along the reporting food chain. But it exists in the research paper itself. A few house plants if left alone in a building will eventually reduce a 5,000 ppm level down to about 200 ppm, at which point photosynthesis stops -- so obviously a time factor must be involved.

In reading the paper, I see their "90% reduction" was primarily achieved through surface adsorption, not photosynthesis -- which means it's a one-time effect. How well this will stack up over the long term compared to natural plants remains to be seen.

An impressive word salad. Devices like the steam engine and electric motors and water pumps have eroded man's ability to dig ditches and plow fields by hand, as well as manually carrying water from the town well back to the hut. Why, housewives of yesteryear used to be strong as oxes, and with good reason. Should we return to those days to avoid losing essential skills to technology?

It actually was adsorption plus photosynthesis, though the volume was small.

Sorry about that. Guess it came out wrong.

But by that logic if you see humanity itself in general being displaced by something superior as an acceptable outcome, then...you are entitled to your opinion.

I think that what was said about eroding skills is entirely valid. Let me put this to an example...and hopefully you can see where I'm coming from.

I work in quality, and one of the core items is an 8D report. It's basically a "what did we screw-up, how did we do it, what are we doing about it, and how do I prove that's actually a solution." If you use chat GPT you can get a 5 whys to be output, what the failure was and why it happened. It takes about a 10 minute exercise and boils it down...with the right inputs...to 30 seconds. You'd think that this was a good thing...until 4 iterations in you finally get something useful...because despite using the broadly trained chat GPT you need to structure your question in pretty much the exact way to get the answer and output format you want. So...instead of doing a thing, you learn and know how to structure a question to an algorithm. Its output isn't different, but the truth of this is that instead of using human skills you are banking on an equation to give you an outcome.

Next week I'm sick. It's easy to grab someone from the floor who's even mildly tech savy. They are, in-fact, better than me because they've spent hours getting LLM "training" by making thousands of former presidents as anthropomorphic cat people queries...and they know exactly how to get exactly what they want out of an LLM. You sit them down in my seat, and tell them "do what he was doing with the AI, and fill out an 8D report." Your LLM master looks back at you and says...what? If they're honest they tell you that garbage in is garbage out, and because they don't understand what to put in they cannot do the 8D fast with the tools at hand (despite thoroughly knowing how to use the tools). If they're dishonest they poke around for an hour or two using the words on the form, getting something, and submitting a "hallucinated" set of LLM results.


That's how I read the "word salad." If that's not fair you can now use Grammarly to process that into something more intelligible with the assistance of AI. In the commercial somebody thinks somebody else is too complicated, and their AI "simplifies" responses...while the other person gets magical assistance in making their response more professional...which largely consists of MS word level word and grammar checks, though this is sold as AI. If it isn't clear...this commercial, something playing on youtube, is why I don't believe that replacing people with LLM computers is a viable thing.

To roll back to other points made, did you know that Urbana didn't actually make concrete better? Do you know what they actually did? I'll give you a hint... (a)*z+(b)*y+(c)*x+...=cement. Yeah...it's that simple. The later half of the alphabet (because I don't remember the keyboard shortcuts for alpha, beta, and gamma) represents the mechanical property (compressive strength) of cement in your concrete, and by adding a bunch together with varying quantities (the front half of the alphabet) you get a bunch of new mixes that might replace cement as a binding agent. It did not spit out something new or valuable...because once the mixes came out you then have to actually test them... It might be something you missed there, but Urbana didn't highlight that there were thousands of compositions. They highlighted a number of novel solutions which needed research. That LLM was incapable of giving any output beyond a mix with compressive strength...which was made infinitely faster than human comprehension only in the matric of iterative calculation...something that the basic calculator proved decades ago. Is my Ti-86 AI powered now? Did they "take our jobs?"

 

[Endymio]

did you know that Urbana didn't actually make concrete better? Do you know what they actually did? I'll give you a hint... (a)*z+(b)*y+(c)*x+...=cement. Yeah...it's that simple....That LLM was incapable of giving any output beyond a mix with compressive strength...which was made infinitely faster than human comprehension only in the matric of iterative calculation...something that the basic calculator proved decades ago.

This couldn't possibly be more wrong. First of all, they didn't use a LLM, but a variational autoencoder. I realize to those who don't work in the field that all of AI is considered to be just the LLM chatbots you see online, but your statement is like claiming the Navy used a fleet of garbage trucks to launch a fighter, instead of an aircraft carrier.

More importantly, they didn't simply iterate out all possible combinations -- there's several million such, far too many to test. Instead, the model, by examining 1,000+ existing concrete recipes, generated an n-dimensional mapping of the relative effects of all input elements and production steps, temperatures, pressures, curing times and other conditions, in formulations that have never before existed. It then estimated which of those formulations best fit the desired performance profile, and output just those few for testing.

Having worked in research for many decades, I can tell you that, while everyone understands a basic linear relationship, the number of people that can even grasp the basic concepts of a multidimensional nonlinear regression is very small, even when the number of variables is only two or three. In this case, there were more than 100 such variables. No human -- with or without "a basic calculator" would ever solve that problem.

I work in quality, and one of the core items is an 8D report ... If you use chat GPT you can get a 5 whys to be output ... That's how I read the "word salad." If that's not fair you can now use Grammarly to process that into something more intelligible

Sure, LLMs of today can't replace a well-motivated, well-trained person. That's today. In five years, an LLM will be better at essentially all tasks than a person of average intelligence, and in 10 years, they'll outperform our best geniuses. I understand how frightening that may be to some. But fear doesn't alter the reality.

 

[R-T-B]

Open a window and have some real plants.

I swear, why are some of you even on a tech site? Some days man.

Hyperbaric chamber therapy is -- barring a few specific conditions -- a baseless fad.

My father was prescribed it to accelerate healing by a real Doctor after his cancerous tumor removal from his neck. It helped. I'll defer to medical scientists here.

 

[Marcus L]

Cut down the trees and replace them with solar farms and artificial CO2 reduction machines, humans really are the dumbest and lowest of life forms, honestly, without CO2 there is no plant life, without plant life there is no food chain, "green" is a scam tax created by the government who have taxed you to walk, drive, eat and now breath, can't wait for the asteroid that ends this virus

 

[R-T-B]

Cut down the trees and replace them with solar farms and artificial CO2 reduction machines

Literally no one said or proposed this, Mr Strawman.

 

[Endymio]

I swear, why are some of you even on a tech site? My father was prescribed it to accelerate healing by a real Doctor after his cancerous tumor removal from his neck. It helped. I'll defer to medical scientists here.

I swear, is reading comprehension a lost art among today's youth? Re-read the portion of your post that I emphasized. Did this "real doctor" suggest your father, his wife, and the whole family begin regularly sleeping in bariatric pods to improve your general health?

 

[Macro Device]

is reading comprehension a lost art among today's youth?

It affects every age group. I'm also facing a whole lot of people who see non-existent things in my writings and can't see what I actually meant. COVID probably severely damages brains, it wasn't that bad five years ago.

Or maybe they need to have their CO2 in check. Who knows. I can only speculate at this point.

 

[R-T-B]

Who rattled your cage? sorry can I not comment without your approval?

I didn't say that either.

I swear, is reading comprehension a lost art among today's youth? Re-read the portion of your post that I emphasized. Did this "real doctor" suggest your father, his wife, and the whole family begin regularly sleeping in bariatric pods to improve your general health?

He suggested it ongoing for a year for my father in regular intervals to promote wound healing, but thank you for clarifying. Your initial post was vague hence the response you are getting.
By the way I am nearly 40.

Or maybe they need to have their CO2 in check. Who knows. I can only speculate at this point.

Social Media is a big factor honestly, but way OT for here.

 

[LabRat 891]

Micronize this, put it into curtains and wall coverings, and I'd be sold. I've been wanting something to safely increase indoor O2 levels, for *years*
Only concern I'd have, would be managing dust from 'clogging' up the cells.

 

[Endymio]

He suggested it ongoing for a year for my father in regular intervals to promote wound healing

Hyperbaric therapy to promote healing of chronic wounds -- especially in the elderly who have problems with angiogenesis -- is one of the few on-label uses of the treatment. If you wind up with a case of CO poisoning or decompression sickness from a diving accident, a reputable doctor will put you in one too. It's use among the general population, though, is a baseless fad.