From Fedscoop:
The Social Security Administration has banned the use of generative artificial intelligence-based tools on agency devices, FedScoop has confirmed. The block, which is temporary, is meant to ensure the security of data and systems.
The agency’s block of these third-party tools was disclosed earlier this month in a management advisory report for this fiscal year, published by the SSA’s inspector general office. The report noted that the decision was made to protect personally identifiable information, along with health, sensitive and other non-public information, that risked being shared through the use of the technology.
The Social Security Administration said the block was a precautionary measure and that the agency has yet to use generative AI. When asked if the ban applies to agency laptops and mobile devices, a spokesperson said the block “is designed to be agency-wide.” ...
11 comments:
I am shocked that an agency that depends on wet signatures and fax machines would block the use of modern technology.
Hand me my abacus so I can figure out this payment.
On the road to Vision 2025 where there will less employees and more technology. This vision has been in the works for years. The dream is to have a seamless agency.
The statement sounds good - truth might be that they are better off not using it because they don't understand how to use is as they might still befunctioning in the technological dark ages... get the abacus out is right...
I suspect this is more than techno-phobia. SSA is always concerned about privacy of sensitive information belonging to the public. Unless SSA develops and maintains its own AI tools, there is too great a risk of exposure -- the AI input and output become property of the AI machine, not the operator. For the same reason, SSA warns employees not to feed PII into Google Translate or other tranlation engines. I hardly think SSA has the capability to create a funcitoning AI.
May as well ban artificial intelligence from the SSA workplace. Human intelligence has long been absent.
Wow, the truth is there are plenty of folks there who could use AI effectively and others who could create AI related functionality for SSA but.... training an AI right now means a huge vacuum that sucks up everything it can to train. SSA data contains PII and any outside AI that was linked to an SSA employee could in theory suck up information's it shouldn't have access to. It's not likely but just the threat means being privacy positive, not technology phobic,
See https://www.scientificamerican.com/podcast/episode/generative-ai-models-are-sucking-data-up-from-all-over-the-internet-yours-included
Before you post something slandering SSA for being behind the curve or something like that, make sure you know the difference between AI and generative AI. SSA has been using AI for over 20 years in some form or another and actually leads the government in some areas of predictive modeling. Things like QDD/CAL, some OQR and OHO functions, and other security protocols use AI. Generative AI is different and there is no reason FO employees or employees in most positions should be using programs like ChatGPT or others in their daily business. Those requests are processed offsite in the cloud and would instantly violate SSA’s privacy rules. Employees are paid to use their brains for generative writing and although I can think of some decent use cases for generative AI, most of it would be useless to SSA because the ins and outs of SSA policy is not incorporated in the learning models these programs use, which increases the likelihood that the generative AI spits out incorrect information.
Even the older and much simpler forms of AI did poor jobs when it came to anything requiring human judgment. Essentially you just got more of the intellectually lazy thinking done on a more massive scale, with a “techie” veneer that made it palatable to the ignorant masses and crusty old people who ate it up. And we don’t even have any idea how this newer generation of AI is working. So this is absolutely the right call.
There are too many examples of AI mistakes. AI can only pull data from what it's fed. If you don't have a good curator of information going in (which takes considerable subject matter expertise) it's going to do poorly. If it learns from biased datasets, you get biased results. AI is good at finding things in common in big datasets, but bad at interpreting context. Finally, it's decision making reasoning is not transparent, so you can't tell what it based an answer on. AI seems poorly suited for anything like disability determination in its current state.
The headline is misleading. AI is not the same as Generative AI. Please understand the difference. SSA uses AI and has for years, if you don’t know this, then you don’t know enough to comment on this thread. Generative AI is a relatively new technology and because SSA holds the personal information for hundreds of millions of people, using it could have serious ramifications. Sometimes ya’ll act like SSA is just a tiny little agency that any change would be of little or no consequence to anyone.
Ha! SSA is against intelligence, artificial or organic. Just call with a question and you will find out!
Post a Comment