Simpler solution : Post adverts for ibogain treatment
So much effort to fight a war on drugs where the a possible cure for the addictive illegal drugs just happens to be the non-addictive illegal drugs.
Drug dealers and dodgy pharmacies illegally touting opioids online – think heroin, fentanyl, codeine, morphine, and so on – may have their collars felt by an AI cop soon. Ish. Maybe. The US Department of Health and Human Services (HHS) has awarded a contract worth $224,864 (£172,000) to S-3 Research, a startup spun out of the …
You also get 20%-30% of the posts that have nothing at all to do with drugs dealing.
If for example, 1% of all posts were related to drug dealing:
For every 1000 suspects it returns, 7 will be actually related to drug dealing, and 300 will be false positives.
If for example, 1 in 1,000,000 posts were related to drug dealing, which is probably closer to reality:
for every 10,000,000 posts return, 7 will be related to drug dealing, and 3,000,000,000 will be false positives. Or to put it another way, 70% accuracy is not meaningfully better than picking posts at random.
Well partly, would be to no longer have US Marines and British forces protecting poppy fields in Afghanistan.
A place where currently, 80-90% of the raw opium is produced.
When the Vietnam war spilled over illegally into the neighbouring countries that were part of the lod Golden triangle, Air Anerica was flying the stuff to Western markets.
“We use natural language processing models to look for words like 'buy,' 'sell,' 'discount,' or if there is a phone number or hyperlink included,"
AI involvement why? Don't see anything there that can't be accomplished by running some basic RegEx against the posts, with regular updates to the dictionary to take into account new terms being used.
Or is it just that governments are unlikely to spunk millions on a basic search function unless it's packaged in the emporer's current latest non-photon blocking fashion?
Maybe if they could have crowbarred blockchain in they'd have even more funding.
That's what they're paying for... Someone who can remember how to regex on a frequent basis instead of the usual
1. Find issue regex can solve
2. Spend hour(s) re-learning regex
3. Test it
5. Spend another hour debugging
7. Rinse, wash and repeat in 6 months.
"If deals are discussed in private messages on Facebook, though, for example, the machine-learning model, which leafs through public posts, will miss all that chatter."
And how exactly does a bot trawling through public posts address that? Note I don't say "address that problem", because the fact that random people trawling through internet comments aren't able to see private messages is in fact the whole point of the word "private", and not a problem that actually needs addressing. I'm sure various government TLAs would disagree on that point, but I don't see how it's relevant given the subject of the article.
Its been used for the last 50 odd years in NA, and it seems that they might not get another 'war in a shithole' this year, so I suppose we should expect the 'war on drugs' to spin back up. Either that or one of the contenders in Libya will come up with six or seven million dollars worth of hardware and spin that one into the next delivery of democracy in the 'Far East'.
(/me rereads that. Holy crap I'm getting horribly cynical and crusty)
Biting the hand that feeds IT © 1998–2020