James Dyson Award – International and Sustainability winners announced | Hacker News Reader

James Dyson Award – International and Sustainability winners announced

jamesdysonaward.org

88 pointsanandvc45 comments

I feel like I keep seeing these as science fair wins and innovation award wins, but where are they in real life? Are they making it to (the biomedical equivalent of) prod?

In the US, the FDA usually has to approve medical devices like this. 23 and Me was paused while they talked to the FDA.

The cloud connection of the award winner is of some concern. Should you include PII and a sample, US health insurers could use that against you in the future for pre-existing conditions.

https://www.fda.gov/medical-devices/vitro-diagnostics/direct...

Isn't it insurance fraud to fail to report a pre-existing condition anyway? Are you suggesting that people engage in fraud?

Consider this: when you initially submitted your data, the NN said you’re fine, and so you forget about it. Later on, the NN is tweaked and would say you actually probably have cancer, but you don’t know because you never bother checking again. The insurance companies get their hands on those results. You’re not committing fraud (because you don’t know), but now the insurance companies are more reluctant to sell you insurance.

(To be fair, I don’t think the insurance companies are allowed to discriminate against you using only that data. They probably need an official, recognized diagnosis from a real human doctor.)

I feel like "innovation award" wins are a strong predictor that you won't see something in real life, as it may indicate that the founders are chasing PR rather than focusing their efforts on building a real product.

I work for a winner of a similar type of award. The tech quite predictably went nowhere. There was an initial surge of interest and funding, all of which was squandered. Primary issues ignored while making more science fair demonstrations with deal-breaker issues of their own. It's been a ride, I'm looking for work.

How successful do you think someone could be by picking up the ashes of the great-idea flareouts, and re-executing them with competent scientific and business practices?

Some of the most simple things are ready to become products, but they're sitting on a shelf. We've had widespre commercial interest, but it seems the boss isn't interested in small things. The more complex and involved ideas have a long way to go, even if there were people dedicated to solving the problems. I doubt a solution exists that will work outside of the lab.

I'm sure we would have been better selling hundreds of cheaper sensors than zero complex ones.

With the usual caveat that a positive test to a rare condition is most likely a false positive.

At least in this case a false positive is unlikely to have any negative effects other than temporary anxiety while awaiting further screening, which is vastly more preferable to a false negative.

The use of COVID PCR tests for "mass testing" of asymptomatic individuals in the UK has resulted in the number of positive "cases" corelating more closely with the number of tests performed than with other measures[1], such as the number of people in hospital or dying with respiratory symptoms. This has been used to justify more extreme lockdown measures.

There have been estimates of the false positive rate[2] but it is likely to vary widely as the tests are administered by untrained personnel and not stored according to the manufacturer's recommendations. Those who test positive in the UK are not usually given confirmatory tests as they are immediately required to isolate, resulting in a large number of current "cases" which may be false positives[3].

[1]https://www.anhinternational.org/news/is-there-a-waning-pand...

[2]https://www.thelancet.com/journals/lanres/article/PIIS2213-2...

[3]https://www.cebm.net/covid-19/pitfalls-of-repeat-testing-ill...

> vastly more preferable to a false negative

depending from the POV.

if a condition has an incidence of 0.001%, production capacity of test kits will probably be geared to that.

if you suddenly have 0.01% false positives, you're outstripping production by a factor of ten

it will also use up production capacity that might be necessary for other kits, driving up both test prices and/or wait time for everyone else.

for example: false positive in Italy are the reason you see a new case slowing down in the last week: https://www.google.com/search?client=firefox-b-d&q=italy+cov...

this is not because the containment measures are working, it's because a lot of tests are redirected to verify the PCR results, eating up test capacity and slowing down for example contact tracing tests, which are competing for the same kits.

so, no. false positives have far more implication than one individual short time of anxiety.

> false positives have far more implication than one individual short time of anxiety.

That was my point - for example in Liverpool 700 "asymptomatic cases" were found (0.4% of those tested) using lateral flow tests[1]. Well below the expected ~2.5% of the population currently expected to be infected, and well within a range which makes it likely that a significant proportion are false positives. All of these people will have been legally required to isolate for two weeks, many likely unable to work.

The problem with mass testing is that the false positive rate is far closer to the actual incidence.

[1]https://www.bbc.co.uk/news/uk-england-merseyside-54966607

Probably, but it's one of those scenarios where the upside of early diagnosis outweighs the inconvenience of false positives, as long as the scores are way better than luck, or WebMD.

For a bunch of stuff that's simply not true.

Over-testing leads to over-diagnosis which leads to over-treatment. People get treatment that causes them harm and which cannot benefit them because they don't have cancer.

I don't think the idea is that someone uses the home test and then goes straight to the hospital for chemo. It's a new test at the beginning of the diagnosis-pipeline, rather than a replacement for the existing tests.

Sure, but there are people in this thread who think that testing can never be harmful and can never put people on a pathway to harm.

PSA testing is similar: you don't go direct to treatment after PSA testing, yet we know that PSA screening does cause over-treatment.

>Sure, but there are people in this thread who think that testing can never be harmful and can never put people on a pathway to harm.

Can you be more specific as to which comment claims that? Given that you're replying to a comment I've made which did not make those claims.

So you send your urine test results to the "cloud". Now the Google engineers will know whether you are going to die even before your doctor does. That's just great.

can't wait to see ads for coffins and start wondering if google knows more than I do.

Reading „cloud“ and „cancer test“ in the same article gives me the shivers. This is arguably one of the most sensitive pieces of information one can own about a person. The NN could just run in the app, recent phones are heavily optimized for this anyway.

On the fence about this. On the one hand I agree that sensitive data should stay local, but on the other hand this would make their very valuable asset (the model) public.

Is it really that scary? In countries with sane laws and/or universal health care you can shout about your cancer diagnosis from the rooftops.

Now imagine you live in a country without sane (or sanely enforced) protective laws _OR_ universal healthcare!

Do cloud tests really change much? How many people will start Googling information about a disease as soon as they have a positive diagnosis, giving Google the information without any cloud tests.

Your argument is that because things are already bad it doesn't hurt to make them worse?

No, it wasn't my argument

I could be googling for a relative, or because of an upcoming test. An uploaded test is a hard fact.

wow, there's no mentioning of Google in the article at all (ctrl-F showed no hit); Even then, "Google engineers" do NOT just get access to GCP client's data... it would be a immense breach of client trust for GCP and would be a code red for google immediately.

Please refrain from making strenuous, baseless connections to arbitrary articles for your own privacy vendetta. It helps nobody.

I used Google as an example of what it normally means to send stuff to the cloud. Sorry If it bothered you. Insert X powerful monopoly private company instead.

HN readers don't seem to be very good at detecting hypotheticals used as narrative devices.

I think the "no jokes or fun" rule hurts comprehension.

How well known is James Dyson in the US?

The reason I ask is that in the UK he is quite a polarising figure, with many people hating on him because he is a Brexiteer with all of his manufacturing now moved out of the UK to the Far East. Plus the products are not liked by all with some daring to say his vacuum cleaners are heavy, expensive and lacking in durability. Others insist on a Dyson product and won't settle for anything less.

Therefore, in the UK, winning an award with his name on it is not going to universally impress people. Plus the prize money is not that great in business terms.

So, how is James Dyson regarded internationally?

>how is James Dyson regarded internationally He's just a guy that makes (good) vaccuums.

And hair dryers

He got some press for that vaporware electric car.

This is really interesting, reading more deeply into it, it sounds like it's been known for a while that there are metabolites that can be used to detect cancer (and the reserach seems to indicate that's true for lots of different cancers), and it seems this person is gaining time to market by simply proving the test works through observation rather than understanding which metabolites are present for which reasons - which is a more traditional approach others are taking. I don't know why they're talking about pepole having their own machine or processing the results in the cloud though. It seems like the underlying test - a few sensors exposed to a urine sample is something you could provide in a thousand different ways (test via mail, routine tests during doctors check ups) without waving yours and going "WooooOOoOOoo Clloooudddd" - although maybe that's the sort of thing you need to do to impress Dyson.

Many in this thread are saying there's no need to upload this information to the cloud because the transform can be performed locally on modern hardware, but that puts significant pressure on the business case. A model in the cloud is protected. Stored locally, that same model is potential shareware. If you're seeking to maximize the number of true positive diagnoses, then a viable business with low marginal costs is a better bet than an open, stagnant research model. Further, as others have noted, this information may already be stored widely in other clouds due to patient post-diagnosis search and browsing activity. Thus, adding another surface is only a linear change to the patient's privacy. Put simply, the incremental privacy change to the end user here may be worth it.

The section about the sustainability winner completely ignores climate change. Mind boggling.

> The problem: Fossil fuels continue to account for over 81% of global energy product according to the International Energy Agency. It is estimated that, if we continue to burn fossil fuels at the current rate, global supplies of gas and oil will deplete by 2060.

Is that relevant? If you do something you are way better than the folks who just virtue signal. The result is what counts. That's what Greta herself says.

Sounds like they picked a scope limited enough to guarantee competing submissions are comparable on the same metrics.

That they didn't prioritize scopes the way you do does not seem that mind boggling to me.

> global supplies of gas and oil will deplete by 2060.

In 1990 I was told fossil fuels will be depleted by 2010.

people are sometimes wrong, but i think the 2010 was sensationalized headlines, not actual science

The prediction was that peak oil (i.e. the maximum rate of oil extraction) would happen in 2000. For conventional oil, this was not far off: production peaked in 2005 and has been in decline since. What the prediction missed is that we would extract shale oil on a large scale. The long-term economic viability of fracking is in doubt, though, since apparently no fracking enterprise has been profitable to date; it's all financed by venture capital.

https://economicsfromthetopdown.com/2020/11/16/peak-oil-neve...

Can anyone explain the security implications of using such a device? Wouldn't companies get access to sensitive health data of a person?

Hard pass on a cancer testing device that stores and processes my data on the company's server or their rented cloud storage.

This could be done locally with current phone or PC tech.

I know "cloud" and "AI" and "machine learning" are Pavlovian signals to venture capitalists, but come on.

Yeah, it needs Hyperconverged Blockchain to “accelerate the vision”.

Reply on news.ycombinator.com