I know this documentary existed, and I thought it was an episode of Frontline. Strangely neither Wikipedia nor even pbs.org has a complete Frontline episode guide, but IMDB does. I went thru it but couldn’t find a match.
The episode aired during the PC revolution (early 90s) and it was about how we were becoming too dependent on computers and software especially. The main reason I remember it is because it contained the first mention I had ever heard of the Therac-25 incident, where a software bug in a radiation therapy machine literally killed people! That incident was still recent enough that the show didn’t even mention “Therac-25” by name, I looked it up years later. It also mentioned how the upcoming version of Windows contained nearly 30 million lines of code. That must have been Windows 95 so it had to be around '93-'95-ish? But like I said I couldn’t find a description of an episode of Frontline from that time (or at all) that matched this. I’m know it was PBS because it didn’t have commercials and I’m pretty sure it was Frontline because I remember Will Lyman narrating it. It wasn’t their episode about Hackers and it wasn’t Cringely’s Triumph of the Nerds either.
Was it just a standalone special? Or did I dream the whole thing…
P.S.
Google [SIZE=“2”]Therac-25, its a chilling story that was successfully hushed up by the company that built it.[/SIZE]
Seem like a perfect match for The Machine that Changed the World, a five-part 1992 PBS documentary series on the history of computers, narrated by Will Lyman. A very small of amount of additional information is available from Wikipedia.
Looks like the entire series is available for online viewing here.
You mean AECL, which is now a case study in every safety-critical systems and bio-engineering course, and which is still paying out multimillion dollar[sup]1[/sup] settlements due to their screwup? Some coverup.
And at 43:50 in the last episode is the part about the Therac-25 (which it turns out they did actually name, I just forgot it). Something else I remember about this was the two different approaches to AI computer scientists were taking back then. One still pursued the idea of essentially creating a thinking engine algorithm, the other instead was simply going to try to create a database of every piece of information possible. Because the internet hadn’t taken off yet this seemed downright crazy then, but now this is exactly what they do! IBM’s Jeopardy-winning Watson machine works this way (although without being connected to the internet). I even saw a recent PBS show with an interview with the guy from this show talking about how much information they’ve accumulated in the last 15+ years.
The Machine that Changed the World, I’ll have to look thru my VHS tapes!
Hey, I’d never heard of it. During the 90s I literally read half a dozen computer mags cover-to-cover and never heard about it. And being taught in college engineering courses and/or paying settlement money does not a headline make. The settlements may have even required hushing it up! Was it big in Canada?
Because you’d never heard of it prior to a PBS documentary you assume it was covered up? If the settlement had some sort of non-disclosure agreement then how do you think this became part of course curricula about safety-critical systems? (Assuming it is, as friedo suggests.)
Here is a lengthy PDF about the incident, which pulls together information from “publicly available sources.” So it definitely seems like if this was a coverup, it wasn’t a very effective one, since there are enough publicly available sources of information out there that some college researcher wrote up a 50 page report on the incident.
Yeah, I guess this is a case of something that I find a lot more interesting than most people, so I’m surprised that more people don’t know about it.
I just watched the end of that last episode. Neat how they start to sorta get a few things right, but miss so much. Agents! Back then, always with the ‘agent’ programs! Even now Suri blows! And that’s the biggest thing they miss, both the immense growth of cellphones and then their merge into being computers. Hell, Star Trek got it right!
For most people I think it would just fall in the larger category of “product liability” in this case it only killed 6 people. There have been some faulty products that have killed thousands. I think the computer angle wasn’t caught by the mainstream because it required too much technical sophistication to really understand it was a software problem.
Plus it was in 1985, if it was in 2005 or 2010 society would have a higher level of familiarity with software in general so that aspect might not have been too hard to understand and thus there may have been greater appeal. But I actually just read the beginning of that PDF and if you only had a 1980s “man on the street” level understanding of software you probably wouldn’t really understand what had happened here…which is probably why it never became very prominent. (Not when car companies make cars that explode and/or engulf motorists in flames and kill several hundred people.)
It’s probably that pdf that I read a while back. I read thru it slowly and carefully. I’m not a professional programmer but have done enough hobby stuff to be able to understand the concepts. In particular the issues of the** incremented flag variable** and the race condition, made me think, eww!, this is so a perfect storm waiting for that just right moment where it will dutifully do something horrible.
To me, what does border on criminal negligence is the Therac-25’s lack of hardware interlocks. Switches and sensors which physically have to be closed for the machine to operate in high-dose mode. Considering the previous model had them (and its software was then used in this model!), and just the nature of the machine itself (a human radiation therapy treatment device having both low and high dose modes) not backing up the software with physical interlocks seems reckless. I assume they were only left out to cut costs? But this thing is inherently a high-ticket item. And they probably made, maybe a few hundred, if even that many? Not millions of them. How much could it have saved?
Many thanks again! Just finished downloading them all (didn’t take nearly as long as I thought) and will now [del]burn[/del] I mean, not burn them to anything…
Heck, this case was discussed (along with that hotel skyway collapse) in a core engineering class I took a few semesters back. I dunno how “hushed up” it could’ve been.