Comment by throwaway111991 | Hacker News Reader

Comment by throwaway111991 | Hacker News Reader

throwaway111991

I was at the CogX artificial intelligence summit in London a couple of weeks ago, and IBM were there in full force.

I made several rounds around all of the stalls, and sat at the bar for a couple of hours with friends, and the whole time I could see the IBM stall, with 4-5 people there, WATSON plastered everywhere and nobody talking to them.

So I went over. I got talking to one of their technical people there,

I am highly experienced in Deep Learning so I started talking about Neural Nets, and he went blank, and admitted he didn't know much about that. I inquired about WATSON's technology and he couldn't answer telling me he didn't know.

I asked about the main use cases, and what makes WATSONs offering better than Deep Learning, he couldn't answer, or even compare on basic levels.

I asked him "What are the coolest uses of WATSON you've seen" and he immediatly went into a canned response about WATSON diagnosing cancer (a project I had seen and was familiar with) we spoke a few minutes on that, and I asked what other cool projects WATSON had been used on ... he had nothing, and I mean literally nothing.

very disappointing

To be honest, if my choice is "I am highly experienced in Deep Learning so I started talking about Neural Nets" and "IBM Watson representative", I start wondering whether it's possible for the whole building to just go up in flames.

Kind of reminds me of these frustrating anti-nuclear advocates that set up a table with some posters at my old college one day. They couldn't give me any real answers why I should support solar wind and hydro over nuclear besides nuclear being "old technology". They had no response to the fact that this state isn't ideal for large scale use of any of the green alternatives or that much more deaths directly result from solar panel installation accidents alone than nuclear as a whole. They couldn't tell me specifically what was supposedly dangerous about current waste storage techniques. The worst part was I googled the organization they were there on behalf of and right on their webpage was an explanation of the failure of nuclear plants in the state to stay profitable (I forget why but it seems really odd since you get so much power from them). I guess they cared more about getting more mileage for less effort dishonestly swaying the local types who are automatically turned off by "radiation" and big industrial buildings.

Was that event not geared towards technical people overall or was that stand an exception?

It sounds like you trolled some kids passionate about a subject to argue for a technology and industry that you don't really understand.

Were you prepared to compare the plight of the solar installers who fall off of roofs and get injured to the construction workers injured on the job during various nuclear plant construction projects?

Did you count the various victims at Chernobyl? (Epidemiologists project at least 4,000 deaths)

Were you ready to justify why states like New York needs to subsidize nuclear plants with $500M in direct savings because they are not economically viable to operate anymore?

In your case, those kids were being nice.

Why "trolled"? Those are all reasonable questions to ask. This advocacy has real world consequences.

It's "trolled" because the OP clearly did not actually understand what he was trying to argue for. He was just going for "gotcha" responses.

If someone is trying to spread awareness of global warming, and someone asks this person "but the climate has changed before, why is it different now", and all the climate change awareness advocate can say is "well, uh, climate change is bad, ok" then it still reflects poorly on the advocate for not being able to defend what they're advocating, not on the person who may have been misinformed but was giving the advocate a chance to show them the error of their ways.

And this isn't even accurate, everything I said is a common pro nuclear response. The only part I wasn't informed on at that moment was the budgeting issue.

I've been the "dumb guy at the table" before.

Not for advocacy, but as a representative of an organization whose policy choices were very controversial for a group of stakeholders. (I cannot talk about specifics, which aren't relevant anyway)

It's not smart or productive to get into a public debate with people in that type of forum. You're there to deliver a message, and engaging in ad hoc public debate is unwise for your cause and disrespectful to your hosts. The only way to win is to not play the game; it's better to just end the conversation in the most tactful way possible (which may include playing dumb or remaining silent) or move it to a private forum.

So what do you do if you're called out for obvious BS?

> Were you prepared to compare the plight of the solar installers who fall off of roofs and get injured to the construction workers injured on the job during various nuclear plant construction projects? > Did you count the various victims at Chernobyl? (Epidemiologists project at least 4,000 deaths)

The Banqiao dam killed 170k, and that's just the largest dam failure, there's a heavy tail following it. Wind and solar don't have any single incident so dramatic, but they require so damn much manufacturing, construction, and maintenance to generate so little energy that they don't do particularly well in the comparisons I've seen.

https://en.wikipedia.org/wiki/Energy_accidents

Admittedly this is a surface level understanding, but you speak with the confidence of someone who knows better while providing even less concrete information. I'm hoping you've got deeper knowledge to back your position up and that you're willing to share.

> Were you ready to justify why states like New York needs to subsidize nuclear...

Everyone in the energy industry gets subsidies. Big subsidies. What matters are costs per kwh, and nuclear seems to win there too.

http://instituteforenergyresearch.org/analysis/electric-gene...

Are these numbers wrong? If they're wrong, and fission is actually much more expensive than it looks, is it because fission is fundamentally hard / risky or because the updated prices factor in risk due to irrational public perception?

I agree that there is plenty of government intervention of energy. But I don't think there are other electrical generating technologies that require direct cash infusions of that scale when in a mature operational state just to remain a going concern.

The costs for fission are high because they must operate at peak capacity to recoup their capital costs. That was fine when electricity generation was 100% regulated and demand curves were steady. Now things are changing faster and we have market pricing for generation. Market forces price peak demand higher than base demand, and smaller, more nimble and cheaper generators are eating the lunch of big 1960s nuclear plants.

The risks and technology issues are also serious and very expensive, but we have pushed out the costs into the future and taxpayers who aren't born yet will be paying those debts. Nuclear advocates always ignore the costs associated with securely storing waste products for a period of time approaching recorded human history. That said, those costs aren't priced into nuclear energy.

If they were so passionate then they should've known more. It was likely just something done as part of a class assignment.

Sure, I don't understand nuclear perfectly - it's their job to convince me of their platform and that involves educating.

Yes, I'd say solar was still more dangerous if it resulted in more deaths per kWh.

The context was the US so I was talking about statistics in the realm of US safety standards. Even worldwide nuclear gets a better score than worldwide everything else: https://en.wikipedia.org/wiki/Energy_accidents

I said in my post that the economics of the operation was a valid point. It was also the only point the students at the table didn't give. My overall point was that it was disappointing how they wouldn't just use the real argument and instead relied on faulty reasoning like "it's old."

Being nice how? If you're putting up a table to convince people of some position, swaying people on the fence is what you signed up for.

> Yes, I'd say solar was still more dangerous if it resulted in more deaths per kWh.

That's a statistic that inherently favors nuclear because:

- It focuses death vs. injury.

- It doesn't normalize for states with tough regulation of falls. Is it reasonable to attribute shoddy labor practices with an energy source? Injuries will be lower in states like New York because employers have 100% liability for falls, and insurance carriers require safety measures.

- It compares installation of solar panels to operational run state for nuclear.

- There are no reliable statistics re: deaths and injuries for construction or operational workers building nuclear plants 40-50 years ago.

An apples to apples comparison that was meaningful would be "Build / Installation Injuries per kWh of Rated Capacity". For operational comparison something like "Injuries per kWh of Generated Capacity".

It's actually a hard comparison that requires some judgement, as operationally a large professionally managed system like a nuclear plant is a low risk, high impact profile for injury and death. Distributed, unmanaged solar generation is likely going to generate more incidents (accidents, transformer fires, etc) but have a lower (and scoped) impact.

> They had no response to the fact that this state isn't ideal for large scale use of any of the green alternatives or that much more deaths directly result from solar panel installation accidents alone than nuclear as a whole.

Given an uninformed student and a misinformed didact, I admit to having a strong bias toward criticizing the latter.

From the Wikipedia page on the human toll of the Chernobyl disaster[1]:

> The [2005 Chernobyl Forum report] says it is impossible to reliably predict the number of fatal cancers arising from the incident as small differences in assumptions can result in large differences in the estimated health costs.

[1] https://en.wikipedia.org/wiki/Chernobyl_disaster#Human_impac...

Chernobyl was a plant design that was already outdated at the time and known to be dangerous, ran in the Soviet Union. It's not representative of anything modernly relevant.

To be fair, the "technical" person at a conference does not mean "engineer". It's like a quarter tier past pure-salesperson. Obviously not an impressive showing overall, but I don't think it should be used as a judge of their technical talent.

I didn't expect a discourse on topology, but Compare your X to Y, and knowing more than 1 use case is basic, basic, sales stuff

Seems like IBM latest sales strategy is I B M and stickers.

I understand that the people sent to a conference may not be highly knowledgeable about some of the companies products. However, if the company isn't going to at least train them for easy questions like why anyone should buy the product featured prominently in the booth's advertisements, then what was the point in renting vendor space?

[1] At a minimum, give them a handbook with the official answers for common questions.

Yea, watson's distinguishing features can't be "magic beans".

A couple of psychologist made an algorithm on punch cards to diagnose stomach cancer in the 1960s more accurately than doctors. Grab a copy of The Undoing Project.

I work in developer advocacy at IBM.

If you would like to follow-up about your experience or to get connected with the right folks to answer you questions, please feel free to reach out to me.

Watson, is that you?

These are parodies, right?

Is it slot filling then to find treatments? I saw the 60minutes episode about how they have it read the literature.

The real problem is that Everybody is afraid to share data with them because they know Watson will become better at the expense of company's business model.

I agree, IBM would never send anyone who actually knows anything to these events. The developers are locked away for several good (for IBM) reasons...

Disclosure: I work in developer advocacy at IBM.

The issue of data privacy and protection is complex. IBM provides explanations and opt-outs in many of the Watson services.

For example, you can set the X-Watson-Learning-Opt-Out header on speech-to-text calls to prevent retention and use of your data for learning. [1]

[1] https://www.ibm.com/watson/developercloud/doc/speech-to-text...

Why not an X-Watson-Retention-Opt-In header?

Are you saying that your model does not improve from learning from customer's data? If the answer is yes, then what prevents you from deploying the model at a customer's competitor?

Reply on news.ycombinator.com