| Sysop: | Amessyroom |
|---|---|
| Location: | Fayetteville, NC |
| Users: | 27 |
| Nodes: | 6 (0 / 6) |
| Uptime: | 46:39:26 |
| Calls: | 632 |
| Calls today: | 3 |
| Files: | 1,187 |
| D/L today: |
24 files (29,813K bytes) |
| Messages: | 176,485 |
https://www.breitbart.com/politics/2025/09/29/uc-berkeley-computer-science-prof-fresh-grads-struggle-to-get-jobs-due-to-ai-a-thinning-of-the-ranks/
Circuit design is hard, which is why 95% of EE grads are CE/EE majors,
people who code but don't draw. And the IC companies slurp up a lot
of the kids that do understand electricity.
I advertised for EE interns and got hundreds of typists, c++ and rust
and verilog coders.
The times they keep a-changin.
John Larkin
Highland Tech Glen Canyon Design Center
Lunatic Fringe Electronics
On 9/30/2025 11:02 AM, john larkin wrote:
https://www.breitbart.com/politics/2025/09/29/uc-berkeley-computer-science-prof-fresh-grads-struggle-to-get-jobs-due-to-ai-a-thinning-of-the-ranks/
Circuit design is hard, which is why 95% of EE grads are CE/EE majors,
people who code but don't draw. And the IC companies slurp up a lot
of the kids that do understand electricity.
I advertised for EE interns and got hundreds of typists, c++ and rust
and verilog coders.
The times they keep a-changin.
John Larkin
Highland Tech Glen Canyon Design Center
Lunatic Fringe Electronics
I'm skeptical of how much heavy lifting AI is doing in practice, I
remember one particular recent experiment I did trying to get GPT to
help with a simple state machine table for detecting key presses from a >"resistor ladder" keypad connected to an ADC, e.g. so you can take a
couple buttons hung off one pin and detect single presses, double
presses, held, and so on.
It eventually came up with something viable but it was painful and
required a lot of corrections and at the end of the day I probably
wasted time on the experiment, I could've just written the table myself
and it would've been better.
It does come in helpful for boilerplate stuff like "I want a range
function in C++ that I can send a value to and it will bin it when it
falls within certain pairs of thresholds and it spits out:
template<typename... Thresholds>
int8_t range_index(uint16_t value, Thresholds... thresholds) {
constexpr size_t N = sizeof...(Thresholds);
static_assert(N >= 2, "At least two thresholds required");
uint16_t arr[] = { static_cast<uint16_t>(thresholds)... };
for (size_t i = 0; i + 1 < N; ++i) {
if (arr[i] <= value && value < arr[i + 1]) {
return static_cast<int8_t>(i);
}
}
return -1; // Not in any range
}
Which works well enough, I probably wouldn't have remembered that it's >possible to directly instantiate a C++ array from a variadic template
like in line 5, the template syntax is monstrous in general.
It's helpful sometimes but it's still kind of a mixed bag and may be a
mixed bag for the forseeable future, I think it's maybe realistically >providing 10-20% real-world productivity boosts not 100-200%.
I think it's more like AI has provided something of an impetus to notice
that a lot of programming jobs are overpaid, I've know people in their
20s making well over 200k in the Boston area doing what seems like not
very much, they spend a few hours a day on coding and then they're in
IRC or on social media the rest of the time.
Which is fine, most humans in the real world are only good for 3-4 hours
of intense focus every day. But IMO these folks are not doing 200k+
worth of work.
On Tue, 30 Sep 2025 15:59:20 -0400, bitrex <user@example.net> wrote:
On 9/30/2025 11:02 AM, john larkin wrote:
https://www.breitbart.com/politics/2025/09/29/uc-berkeley-computer-science-prof-fresh-grads-struggle-to-get-jobs-due-to-ai-a-thinning-of-the-ranks/
Circuit design is hard, which is why 95% of EE grads are CE/EE majors,
people who code but don't draw. And the IC companies slurp up a lot
of the kids that do understand electricity.
I advertised for EE interns and got hundreds of typists, c++ and rust
and verilog coders.
The times they keep a-changin.
John Larkin
Highland Tech Glen Canyon Design Center
Lunatic Fringe Electronics
I'm skeptical of how much heavy lifting AI is doing in practice, I
remember one particular recent experiment I did trying to get GPT to
help with a simple state machine table for detecting key presses from a
"resistor ladder" keypad connected to an ADC, e.g. so you can take a
couple buttons hung off one pin and detect single presses, double
presses, held, and so on.
It eventually came up with something viable but it was painful and
required a lot of corrections and at the end of the day I probably
wasted time on the experiment, I could've just written the table myself
and it would've been better.
It does come in helpful for boilerplate stuff like "I want a range
function in C++ that I can send a value to and it will bin it when it
falls within certain pairs of thresholds and it spits out:
template<typename... Thresholds>
int8_t range_index(uint16_t value, Thresholds... thresholds) {
constexpr size_t N = sizeof...(Thresholds);
static_assert(N >= 2, "At least two thresholds required");
uint16_t arr[] = { static_cast<uint16_t>(thresholds)... };
for (size_t i = 0; i + 1 < N; ++i) {
if (arr[i] <= value && value < arr[i + 1]) {
return static_cast<int8_t>(i);
}
}
return -1; // Not in any range
}
What a hideous mess.
Some day there will be a nice way, not necessarily "AI", where
ordinary people can describe what they want in a clean, visible form,
and the computer will just do it. Soon, I hope.
On Tue, 30 Sep 2025 15:59:20 -0400, bitrex <user@example.net> wrote:
On 9/30/2025 11:02 AM, john larkin wrote:
Some day there will be a nice way, not necessarily "AI", where
ordinary people can describe what they want in a clean, visible form,
and the computer will just do it. Soon, I hope.
On 9/30/2025 5:39 PM, john larkin wrote:
On Tue, 30 Sep 2025 15:59:20 -0400, bitrex <user@example.net> wrote:
On 9/30/2025 11:02 AM, john larkin wrote:
https://www.breitbart.com/politics/2025/09/29/uc-berkeley-computer-science-prof-fresh-grads-struggle-to-get-jobs-due-to-ai-a-thinning-of-the-ranks/
Circuit design is hard, which is why 95% of EE grads are CE/EE majors, >>>> people who code but don't draw. And the IC companies slurp up a lot
of the kids that do understand electricity.
I advertised for EE interns and got hundreds of typists, c++ and rust
and verilog coders.
The times they keep a-changin.
John Larkin
Highland Tech Glen Canyon Design Center
Lunatic Fringe Electronics
I'm skeptical of how much heavy lifting AI is doing in practice, I
remember one particular recent experiment I did trying to get GPT to
help with a simple state machine table for detecting key presses from a
"resistor ladder" keypad connected to an ADC, e.g. so you can take a
couple buttons hung off one pin and detect single presses, double
presses, held, and so on.
It eventually came up with something viable but it was painful and
required a lot of corrections and at the end of the day I probably
wasted time on the experiment, I could've just written the table myself
and it would've been better.
It does come in helpful for boilerplate stuff like "I want a range
function in C++ that I can send a value to and it will bin it when it
falls within certain pairs of thresholds and it spits out:
template<typename... Thresholds>
int8_t range_index(uint16_t value, Thresholds... thresholds) {
constexpr size_t N = sizeof...(Thresholds);
static_assert(N >= 2, "At least two thresholds required");
uint16_t arr[] = { static_cast<uint16_t>(thresholds)... };
for (size_t i = 0; i + 1 < N; ++i) {
if (arr[i] <= value && value < arr[i + 1]) {
return static_cast<int8_t>(i);
}
}
return -1; // Not in any range
}
What a hideous mess.
Some day there will be a nice way, not necessarily "AI", where
ordinary people can describe what they want in a clean, visible form,
and the computer will just do it. Soon, I hope.
C++ is for expressiveness and generating assembly that's efficient
and/or compact object file size, not really beauty.
Ask it to generate some Python, that's often pretty close to pseudocode
I advertised for EE interns and got hundreds of typists, c++ and rustI don't recall you advertising here. My young one spent the summer at Meta but would have preferred an analog hardware intern experience.
and verilog coders.
On 9/30/2025 8:02 AM, john larkin wrote:
I don't recall you advertising here. My young one spent the summer at Meta but would have preferred an analog hardware intern experience.
I advertised for EE interns and got hundreds of typists, c++ and rust
and verilog coders.
My young one spent the summer at Meta but
would have preferred an analog hardware intern experience.
[A friend asked me to talk to his son as he was declaring a major.
He'd seen how "free" my lifestyle was and tried to point that
out to his kid.-a Kid was looking into psychology.-a (Really?
Have you checked to see if you'll be able to afford your own
APPARTMENT on your starting salary?)
I made no recommendations
other than "make sure you are going to ENJOY this cuz you are
likely going to be doing it for 50 years!"-a He ended up getting
interested in software and his starting salary (15 years ago)
was $95K.-a His Dad was grateful to me that he had rethought the
psychology option!]
Is that his intended career path?-a Or, just an "interest"?
On 10/1/2025 4:02 PM, Don Y wrote:
Is that his intended career path?-a Or, just an "interest"?
Last year the six months internship with Apple was doing board level work and
was enjoyable. The Meta hardware engineering internship turned out to be writing a lot of Python in support of a larger hardware goal which wasn't what
was hoped for.