Ryan, this is great, I came here for an argument! Thanks. ("I wish to believe 'snow is blue' if and only if snow is blue")

(a) OK, can we agree on "In most cases with 'normal' 0.3<TSH<5.5, TSH is the sole test performed"?

(b1) I don't know, but given that FMS includes 'brain fog' and 'tiredness' I'd be surprised if many people with it haven't had a TSH test. I would be surprised by the existence of people who only have the tender points and no other hypometabolic symptoms. Do we know what proportion that is?

(b2) "Also there are studies where they stimulate release of thyroid humans. There are literally hundreds of highly relevant studies here." I don't quite understand what you're saying here. Can you link to a couple? Google scholar "fibromyalgia and thyroid" gives top hit: http://europepmc.org/abstract/med/1512769, Neeck G , Riedel W "Thyroid function in patients with fibromyalgia syndrome.", in which they find abnormalities in a thyroid hormone releasing test in fibromyalgia patients. Doesn't that support me?

(c) There are forms of hypothyroidism that don't show up on the TSH test certainly, 'central hypothyroidism' and 'peripheral resistance to thyroid hormone', which have the same presentation but normal TSH. 'Central' should give you normal TSH but low T4 and T3. 'Peripheral' should be normal in all respects. But they're thought to be vanishingly rare, and as far as I know, CFS/FMS people aren't tested for them. In fact presumably the only way to test for them would be a trial of thyroid hormones! That's kind of my point.

(d) Not just tiredness and slowness. It's more of a general metabolic collapse. And which systems fail first seems to be random, which is why it's so difficult to diagnose clinically.

(e) I'd expect anyone with widespread pain to recoil if you tried to touch them.

"What is needed here are similarities that are specific to these conditions and that are shared, whereas here it is the aspects that are the most general that are shared but the main facets of each condition are quite different and overall the two conditions don't blur together in an interesting way."

John Lowe appears to have known the rheumatology and endocrinology literature backwards, and claimed that every symptom of fibromyalgia was a symptom of hypothyroidism and vice versa.

He gives references for a vast number of them in:

Inadequate Thyroid Hormone Regulation as the Main Mechanism of Fibromyalgia: A Review of the Evidence

You can find it here: http://tpauk.com/main/dr-john-lowe-overview-of-important-articles/

There were a couple of hypo symptoms (low basal temperature and low basal metabolic rate), which are arguably the characteristic symptoms of the disease, that hadn't been documented in fibromyalgia. So he got a bunch of FMS patients and checked. And found them.

I really think that at this point, someone should be running a PCRT on 'giving desiccated thyroid to FMS patients'! I would actually be surprised if T4 alone didn't have some effect, but since Lowe thought (from extensive experience, but based on a sample that must have been skewed by people going to him after doctors failed to help) it was the least useful of all the thryoid hormone therapies, we should take him at his word and try desiccated thyroid. My own bet would be that T4/T3 in the same combination as secreted by the thyroid gland would be the best thing, but that's just a detail.

Again, many thanks! I don't want to have these beliefs if they're false. Take me down.

Hi Ryan, thanks. My point is that 'hypothyroidism' is defined by high TSH. I wonder if it's possible to have 'hypometabolism' or 'all the symptoms of hypothyroidism' whilst having a normal or even low TSH and maybe normal levels of the blood hormones too.

to answer your points:

(a) Yes, agreed. Only wrinkle here is that labs often refuse to check T4 if TSH is normal. This has happened to me twice despite my doctor specifically requesting an FT4 test. And I personally know someone who had CFS for years and lots of expensive NHS and BUPA and privately paid treatment without any thyroid test except for TSH.

(b) Yes, but if you've already removed all the people with strange hormone levels (because they're hypothyroid, and so they don't have CFS or FMS), then of course you'll find normal levels in the remaining cases.

(c) Agreed, because hypothyroidism is defined by TSH. My hypothesis is exactly that many cases of FMS/CFS are 'something wrong with the thyroid system that doesn't show up on laboratory hormone level tests'.

(d) Hypothyroidism has lots of famously non-specific symptoms. Do hypothyroid patients sometimes complain of pains in the same way?

(e) that's interesting and surprises me. I was under the impression that exactly that kind of widespread pain was one of the (many possible) symptoms of hypothyroidism. I'll go and look.

Thanks ever so! I've added the explicit hypothesis to the summary. I'll go and see what I can find out about pain in hypothyroidism/fibromyalgia.

I've only skimmed this but let me give a bunch of info that seems relevant. (As I understand it, the question is whether low thyroid levels could cause chronic fatigue and fibromyalgia). For background, I am a junior doctor with no specific experience in endocrinology and support healthy scepticism towards the medical establishment:

• Thyroid hormones are of course the first thing that one thinks of in people with low mood. Sometimes one will check the T4 and T3 as well as the TSH.
• it is also one of the first things that one would research and it looks like there are plenty of papers checking hormone levels in these conditions.
• People with hypothyroidism usually have high TSH. It would be surprising if this was different for people with fibromyalgia unless they had a specific insensitivity to thyroid hormone, which seems like a different hypothesis.
• people with hypothyroidism usually don't complain of pains in the same way as people with fibromyalgia.
• people with fibromyalgia usually have a high sensitivity to slight painful stimuli in a way that I've never seen from people with hypothyroidism.

I also must admit that the since the post seems to me to be not just wrong but also very slow-moving, I haven't read all of it. I would suggest rewriting this in a few paragraphs with citations to get feedback on a more brief version.

I've only skimmed this but let me give a bunch of info that seems relevant. (As I understand it, the question is whether low thyroid levels could cause chronic fatigue and fibromyalgia). For background, I am a junior doctor with no specific experience in endocrinology and support healthy scepticism towards the medical establishment:

• Thyroid hormones are of course the first thing that one thinks of in people with low mood. Sometimes one will check the T4 and T3 as well as the TSH.
• it is also one of the first things that one would research and it looks like there are plenty of papers checking hormone levels in these conditions.
• People with hypothyroidism usually have high TSH. It would be surprising if this was different for people with fibromyalgia unless they had a specific insensitivity to thyroid hormone, which seems like a different hypothesis.
• people with hypothyroidism usually don't complain of pains in the same way as people with fibromyalgia.
• people with fibromyalgia usually have a high sensitivity to slight painful stimuli in a way that I've never seen from people with hypothyroidism.

I also must admit that the since the post seems to me to be not just wrong but also very slow-moving, I haven't read all of it. I would suggest rewriting this in a few paragraphs with citations to get feedback on a more brief version.

The Facebook group is closed. Should people here assume that they will be allowed to join?

Is the end-game to do data-analysis on data for charity evaluation, intervention evaluation, cost-effectiveness and that kind of thing?

Or, to inform people interested in machine learning about AI safety?

What impact has ml4g had thus far?

Machine Learning for Good is A machine learning and deep learning study group for EAs and rationalists that I'm facilitating.

It includes a study group for the current Udacity Tensorflow/Deep Learning course. I'm not going to repost further info here, one can access it through the following group:

https://m.facebook.com/profile.php?id=1582428355359588&tsid=0.6936991019174457&source=typeahead

I'd summarise the results of your post, just to check my understanding and to allow myself to be corrected.

The utility function of a value-learning agent will change as it encounters evidence, and depending on its actions.

This is a good ideal yet we should note two traps with this approach:

1) if an agent uses its current utility function to evaluate future actions, then it will avoid learning a different utility function - because a change in utility function will cause the agent to behave poorly based on its current standard (i.e. goal preservation from Omohundro's AI Drives)

2) So suppose that the agent uses its future utility function to evaluate its future decisions. Then, in a bad case, it will choose to learn the utility function that is the easiest. The bad case is that the agent thinks that certain actions will change its utility function in a predictable way. i.e. if E[p(U=u1 | a)] != p(U=u1). So we must enforce this axiom of probability E[p(U=u1 | a)] == p(U=u1) to prevent the agent from assigning itself an easy utility function.

I'd summarise the results of your post, just to check my understanding and to allow myself to be corrected.

The utility function of a value-learning agent will change as it encounters evidence, and depending on its actions.

This is a good ideal yet we should note two traps with this approach:

1) if an agent uses its current utility function to evaluate future actions, then it will avoid learning a different utility function - because a change in utility function will cause the agent to behave poorly based on its current standard (i.e. goal preservation from Omohundro's AI Drives)

2) So suppose that the agent uses its future utility function to evaluate its future decisions. Then, in a bad case, it will choose to learn the utility function that is the easiest. The bad case is that the agent thinks that certain actions will change its utility function in a predictable way. i.e. if E[p(U=u1 | a)] != p(U=u1). So we must enforce this axiom of probability E[p(U=u1 | a)] == p(U=u1) to prevent the agent from assigning itself an easy utility function.