College counseling and AI

[Anonymous]

Anonymous wrote:

Anonymous wrote:Only people who don't understand what AI is would trust its advice.

this sounds like college counselor speaking

Yeah, it was.

[Anonymous]

AI has enormous value, but you need to know how/when to use it.

What I think has less and less utility (especially for selective colleges) is those old school regional IEC counselors who create a Google doc and show kids how to do school research, help them organize their college lists (and create a list), and organize ECs. They give minimal strategic advice and minimal strategic essay guidance.

I think good, high-quality essay feedback/editing and highly tailored selective college advice (e.g., here are the 3 T20 schools expanding funding into DC's major that you should look at because the department/field is now those schools' institutional priority) is much more valuable. From what I can tell, fewer traditional counselors are doing this kind of highly bespoke and targeted counseling on an ad hoc basis than people think. It's the only real value in private college consulting now. Some of the big shops do it, but you have to pay for their ultra-premium pricing.

P.S. I am a freelance counselor doing this type of work now; the # of people coming to me after paying the old-school regional IEC anywhere from $10-20k and now urgently need strategic help in RD is more than I can handle. I don't advertise (and won't ever here), word of mouth only. I have posted before when someone was trying to decide on getting into the industry. I had a multi-decade high-profile professional career, retired, and then saw the strategic gaps in this entire process.

Parents: choose your counselors wisely. Use AI where it makes sense (judiciously) and get smart on the process.

[Anonymous]

Yes. AI can be an incredible help if you know how to use it properly. Everything from brainstorming essays to editing specific paragraphs or sentences to helping hit word limits.

As with everything - including a paid college counselor with excellent references and a strong track record - the key is to encourage your DC to put in the necessary work, be assertively authentic, and develop/use their critical thinking skills.

AI is a tool. With all due respect, so is a paid college counselor. Tools can be helpful or they can useless or even damaging. It's all in how you employ them.

[Anonymous]

Anonymous wrote:

Anonymous wrote:AI is an echo chamber. If a substantial number of sources start putting out articles saying something wrong but new, such as "extensive research by leading medical researchers shows that tea leaves cause autism in children," and a large number of popular websites agree to repeat the wrong claim, after a while, AI would repeat the same wrong claim. AI is a parrot, that fortunately is often right because what's out there on the web is often right. But AI can be fooled with a concentrated effort to train it with garbage.

Kinda. The thing is, though, that companies training foundational models are aware of this. The models aren’t just indiscriminately ingesting stuff. Microsoft tried that with Tay, which continuously learned from interactions with users. Twitter taught it to be a Nazi misogynist within 24 hours. Now companies curate what models ingest more carefully - well, models other than Grok.

There’s still a massive “garbage in, garbage out” problem. The PP who’s so taken with Claude still had to feed in a ton of data for it to process. Choose the wrong data (train it on DCUM, for example), and the results will be worthless.

[Anonymous]

Anonymous wrote:

Anonymous wrote:AI is an echo chamber. If a substantial number of sources start putting out articles saying something wrong but new, such as "extensive research by leading medical researchers shows that tea leaves cause autism in children," and a large number of popular websites agree to repeat the wrong claim, after a while, AI would repeat the same wrong claim. AI is a parrot, that fortunately is often right because what's out there on the web is often right. But AI can be fooled with a concentrated effort to train it with garbage.

Kinda. The thing is, though, that companies training foundational models are aware of this. The models aren’t just indiscriminately ingesting stuff. Microsoft tried that with Tay, which continuously learned from interactions with users. Twitter taught it to be a Nazi misogynist within 24 hours. Now companies curate what models ingest more carefully - well, models other than Grok.

PP. For critical issues that can affect the society and the world, such as misleading AI into behaving like a Nazi misogynist, a racist, a right/left-wing extremist, or someone who encourages users to commit suicide, surely the companies would step in and put a stop on it. But our everyday lives is filled with a gazillion non-critical information and issues, many of which are inherently uncertain/subjective, are filled with nuances, either just happen or have yet to happen or can change over time, that AI often screws up. It's simply not feasible for companies to check everything. AI is "use it with your discretion" and that is unlikely to change anytime soon.

[Anonymous]

Smart parents definitely won’t waste money on private counseling, but there aren’t lack of dumb ones who are willing to pay to reduce their anxiety.