I can’t say I’ve ever been a fan of ABC’s Q&A. Members of the panel seem to be selected for entertainment value rather than whether they actually know much about the topics under discussion. In September last year, Q&A featured Ted O’Brien, Shadow Minister for Climate and Energy, Will Shakel, a 17 year old advocate for nuclear energy in Australia, Chris Bowen, Minister for Climate Change and Energy, Nickey Hutley, described as an Independent Economist (amazing how a words meaning changes when you capitalise that first letter … Independent Economist, Liberal Party, Labor Party), and Allegra Spender, Independent MP for Wentworth … aka Teal.
The topic of nuclear waste was touched upon. Pretty clearly nobody except the 17 year old knew much about the issue, so they had to resort to political arguments. Hutley used an argument which went roughly as follows: “Plenty of people are afraid of it, therefore people will always be afraid of it, therefore …”. She could derive anything from these kinds of premises. Indeed people who know nothing about nuclear waste, and have been told by other people who know nothing about it that it poses some kind of (always unspecified) threat to their children's children and the environment in general for hundreds of thousands of year are apt to be understandably fearful. And the level and nature of discussion on forums like Q&A won’t help because it is guaranteed to be superficial.
Can ChatGPT help?
ChatGPT on nuclear waste
Nuclear waste hasn’t been high on the agenda of topics discussed following Peter Dutton’s nuclear announcement. But it’s always hanging around in the background and assumed to be some kind or unsolved problem. It isn’t. Isn’t what? Isn’t much of a problem and what there is was solved decades ago, assuming you decide not to keep and use the incredibly valuable stuff that is called nuclear waste.
I always use the waste issue as a litmus test. You can tell whether people know anything at all about nuclear power by checking if they are worried about waste. If they are worried, then their knowledge, if any, is incredibly superficial.
I can illustrate this by having a conversation with ChatGPT.
Ask ChatGPT a general question and it will predict an answer based on the stuff it has been trained with. It will be superficial because it is predicting answers and answers to simple questions about nuclear waste are all over the web and typically trite.
But dig deeper and AI will also have to dig deeper. And unlike somebody on Q&A, limited by their pre-existing knowledge, AI will be able to dig deeper.
Eventually, assuming you ask the right questions, ChatGPT should hit bedrock: experts and facts. The trick is to keep asking for facts in a way that filters out opinions.
There are two great things about doing this with ChatGPT rather than an actual expert:
Real experts are hard to come by and if you aren’t one you can easily get fooled by pretend experts, people like Chris Bowen. He has the politician’s skill of coming up with things that look like facts and can easily fool you into thinking he knows something when he doesn’t. On Q&A this was evidenced when he claimed that SMRs produce “upto 30 times” more waste than big reactors. If Q&A had been able to quiz him about this, he’d have come up short. Actual experts making claims like that typically deliver them with a bunch of caveats. In this case the caveats are irrelevant. The correct response to that rubbish was “So what?”. 30 times more than a trivial amount is still a trivial amount and if it’s useful, which nuclear waste is, then having more of it is terrific!
Real experts can have incredibly poor communication skills. Before ChatGPT, you had to repeatedly ask them to clarify what the hell they were talking about.
Let’s give an example. I’ll probe ChatGPT on nuclear waste. I’ll abbreviate the answers, they can be lengthy, but you can find the full unedited transcript here.
Me:
What are the risks associated with nuclear waste?
ChapGPT:
Nuclear waste, a byproduct of nuclear reactors, fuel processing plants, and nuclear medicine, poses several risks that need careful management. Here are some of the key risks associated with nuclear waste:
Radiation Exposure: Nuclear waste contains radioactive materials that emit ionizing radiation, which can be harmful or fatal to living organisms. Prolonged exposure to high levels of radiation can cause acute radiation sickness, cancer, and genetic damage. … Long-Term Waste Management: Some radioactive materials remain hazardous for thousands or even millions of years. Ensuring safe storage and containment over such long periods is a significant technical and logistical challenge.
Wow. At this level, ChatGPT is summarising at a superficial level. It’s just like a Q&A guest who is repeating stuff they heard somebody say somewhere and it sounded okay. The AI has found plenty of claims about waste being dangerous for millions of years on the internet and is repeating and summarising them.
Now let's dig deeper. Giving ChatGPT an actual time span will focus its responses.
Me:
How can nuclear waste buried underground for 1000 years harm anybody?
ChatGPT
nuclear waste buried underground for 1,000 years may seem secure, it can still pose significant risks due to a variety of factors: Long Half-Lives: Some radioactive isotopes in nuclear waste have extremely long half-lives, meaning they remain radioactive for thousands to millions of years. Even after 1,000 years, these isotopes can still pose a risk if they migrate to the surface or into groundwater.
At this stage it is still summarising people who don’t know much. How can I tell. It’s not quantifying the risks. Aircraft can crash. That’s a risk, but people accept it because they have a working understanding of the size of the risk; even if they don’t know the numbers. The risk is incredibly small. Some risks are both small in size and trivial in consequences. The risk of a paper cut is one of these. The AI hasn’t get given me anything that tells me the kind and size of risk posed by radioactive waste. The earlier response mentioned that radiation could be fatal. But can the doses of radiation you’d get from waste after a long time in the ground be fatal?
To force ChatGPT to summarise experts who know the size and nature of the risks we need to push it harder.
Me:
What dose of radiation per hour could nuclear waste deliver after 1000 years?
[Note how this question asked for a number; a quantified answer. This usually separates out the actual experts from those who merely spin a good yarn.]
ChatGPT:
The specific dose of radiation per hour that nuclear waste could deliver after 1,000 years depends on several factors, including the type and amount of radioactive isotopes present, the effectiveness of the containment materials, and the distance from the waste.
Key factors include :
Type and Half-Life of Isotopes: Different radioactive isotopes decay at different rates, known as their half-lives. After 1,000 years, isotopes with long half-lives (such as plutonium-239 with a half-life of 24,100 years) may still be quite radioactive. However, isotopes with shorter half-lives (such as cesium-137 with a half-life of 30 years) will have decayed significantly by that time.
Decay and Reduction Over Time: As radioactive isotopes decay, their radiation levels decrease. The rate of decay determines how much radiation is left after 1,000 years.
Now we are getting into expert territory. The answers describe various elements and their properties. People pretending to be experts, will sometimes mention Plutonium-239, and its long half life, in an effort to appear to know stuff. This is the Chris Bowen strategy.
We need one more question to finally get to the motherload.
Me:
What dose per hour could 1 kg of plutonium-239 deliver after 1,000 years?
ChatGPT:
To determine the radiation dose rate from 1 kg of plutonium-239 after 1,000 years, we need to understand several key points: … [lots of detail]
Activity: The activity of a radioactive substance is measured in becquerels (Bq), which indicates the number of decays per second. For Pu-239, the specific activity is around 2.3 x 10^9 Bq per kg.
Radiation Emitted: Pu-239 primarily emits alpha particles. Alpha particles are heavily ionizing but have very low penetration; they can be stopped by a few centimeters of air or a sheet of paper. However, if the alpha-emitting material is inhaled or ingested, it can be extremely hazardous.
Dose Calculation: The dose rate from external exposure to alpha emitters like Pu-239 is generally low unless the material is inside the body. For external exposure, the primary concern would be gamma radiation, which Pu-239 emits in much smaller quantities. …[more details]
Summary: After 1,000 years, 1 kg of plutonium-239 would still be highly radioactive due to its long half-life. However, the external exposure dose rate from unshielded Pu-239 is low because alpha particles have very limited range in air. The primary hazard from Pu-239 is if particles are inhaled or ingested, leading to severe internal contamination and radiation dose.
Finally. We have hit the motherload. A detailed discussion of radiation. Radiation is the stuff that is supposed to be dangerous. It’s the stuff of fear and legend. Most people are afraid of it, but they don't know why. More importantly, they don’t know the how of it.
For example, there is a forest near the Chernobyl power plant that is still massively radioactive. But you can walk around it quite safely. You can lay on dirt laced with plutonium, again, without risk. Why? ChatGPT has explained it by calling on expertise from somewhere on the web.
ChatGPT has even been kind enough to provide a summary of the detail. And I’m going to summarise the summary.
After 1,000 years, the only way nuclear waste can hurt you is if you dig it up and swallow it.
This makes it the same as many other things you might dig up and swallow, or buy at Bunnings and store in your garage, or under a sink.
Humans have been building buildings which can stand for a thousand years for well over a thousand years. In contrast, keeping things buried for a thousand years is much easier.
I also asked ChatGPT for alternative uses for nuclear waste. There are many. It’s only waste if you don’t use it … but that’s grist for another mill. Will Shakel knew this. But then again, he’s 17. Ask him in 30 years when he’s spent decades fighting his way up the ladder in some political party, like Chris Bowen. His mind will be so full of political trivia about process and factions and all the other things that occupy political minds, that he’ll probably have forgotten!
In any event, the bottom line here is that nuclear waste isn’t a problem and hasn’t been one for decades. The technology for vitrifying it is well understood, after which time it’s amazingly robust. The principle problem is misinformation from people masquerading as experts but whose knowledge is far shallower than that of ChatGPT (or Will Shakel).
I used to work around radiation; well done! (& humorous)