Eliza

1.7K posts

Eliza banner
Eliza

Eliza

@eliza_doesstuff

Ex-Physicist, ex-Uber, ex-Data Scientist. The world is big.

Mediterranean, long ago NYC Katılım Mart 2012
520 Takip Edilen165 Takipçiler
Eliza retweetledi
Norman López
Norman López@LopezNorman44·
❌el tutuki splash ✅El Llobregat al seu pas per Sallent
Català
24
381
4.1K
77.8K
Eliza
Eliza@eliza_doesstuff·
@paulg @nntaleb I am not entirely sure the photography is at fault. It also became abstract and “weird” just like paintings. Literally tried to experiment with form similarly, esp poesy - just harder there. Maybe it was just a response to times overall.
English
0
0
0
359
Paul Graham
Paul Graham@paulg·
@nntaleb It may not be so bad with literature. The reason painting declined was that it became possible to win by having a distinctive visual brand (rather than painting well). I doubt there's any analogue in literature.
English
30
4
177
42.4K
Nassim Nicholas Taleb
Nassim Nicholas Taleb@nntaleb·
Until photography, painters competed largely on the verisimilitude of representation. Photography destroyed art by making artists switch to absurd images often driven by social contagion. Now imagine what AI will do to literature.
English
292
326
3.5K
512.5K
Eliza retweetledi
Jaana Dogan ヤナ ドガン
Do you remember the times when the biggest controversy was spaces vs tabs?
English
35
20
355
30.4K
Eliza
Eliza@eliza_doesstuff·
@veldtclix I don't think being bad is what will stop them from replacing us
English
0
0
0
4
Veldt
Veldt@veldtclix·
@eliza_doesstuff they are supposed to replace us lol (don't break my bubble)
English
1
0
1
42
Eliza
Eliza@eliza_doesstuff·
How in this day Google Translate can be so bad as to offer a false friend as a translation?? What the hell did they train their models on? It's not even the first time this happens
Eliza tweet media
English
1
0
1
63
Eliza
Eliza@eliza_doesstuff·
@daniel_nguyenx Afaik Garmin uses a simple formula that connects speed with heart rate and some other devices add age and other “features”, so their measurements are better population averaged, but off on an individual level
English
0
0
0
43
Daniel Nguyen
Daniel Nguyen@daniel_nguyenx·
I’ve been using a Miband 10 to track VO2 Max and it turns out to be wildly inaccurate 😭 Switched to a Garmin and it reports +7 points instead.
Daniel Nguyen tweet media
English
3
0
7
1.8K
Eliza
Eliza@eliza_doesstuff·
@thorstenball JIRA has already took advantage of that. Every day, I see the login popup, login with google popup AND an ALERT that jira wants you to login. At the same time!
English
0
0
0
70
Thorsten Ball
Thorsten Ball@thorstenball·
Good news! There's still some space on the left. Something could show another tooltip or popup or modal there.
Thorsten Ball tweet media
English
4
0
26
4.9K
Eliza
Eliza@eliza_doesstuff·
@GergelyOrosz Same for the internal tech support in a company I work for :/ The thing is these bots don’t have to be awful, but they are worse than AI search on Confluence somehow!
English
0
0
0
16
Gergely Orosz
Gergely Orosz@GergelyOrosz·
Amazon used to be the pinnacle for customer support - but now I have to go thru a useless AI bot to connect to a human to fix my problem. Never thought I'd be this grateful to be talking to a human - these AI support bots are just truly terrible, even for Amazon
Gergely Orosz tweet media
English
22
18
195
17.6K
Eliza
Eliza@eliza_doesstuff·
@GergelyOrosz This is true in my experience. Our company allowed AI and the first candidates I saw either would get stuck or not based on their experience. AI printed code but couldn’t save.
English
0
0
0
80
Gergely Orosz
Gergely Orosz@GergelyOrosz·
Btw this CTO said this interview type works very well. As he told me: "We get the same signal as pre-AI. That is good engineers who know how to build pass, those 'vibe coding' get stuck. The dev who solved this fastest and best actually didn't use AI - was just a machine"
English
8
11
562
111.4K
Gergely Orosz
Gergely Orosz@GergelyOrosz·
From a CTO at a startup: "We interview devs by giving them a task to build an app on the spot, from scratch (2x BE endpoints, some frontend.) They can use AI, ofc - and we dig into why they did this or that. What is surprising: 14/15 devs from Meta failed this screening."
English
312
202
4.8K
2.1M
Eliza
Eliza@eliza_doesstuff·
@GergelyOrosz It just really wanted to be helpful! If there are few duplicates, how can it helpful?
English
0
0
0
65
Gergely Orosz
Gergely Orosz@GergelyOrosz·
A recent example: I pasted very long text into Claude, and asked it to identify duplicate parts that can be removed, showing exact quotes. It hallucinated parts, with quotes, that do not even exist in any input!
English
4
1
29
7.5K
Gergely Orosz
Gergely Orosz@GergelyOrosz·
Every single time an LLM hallucinates, I am grateful: Grateful that I spotted it, and thus remind myself that any and all LLM output needs to be validated. You can never trust these things 100%, unless you have additional validation in place that is 100% reliable.
English
12
10
157
13.1K
Eliza
Eliza@eliza_doesstuff·
@pixperk 🐭 What is it about?
English
1
0
2
448
yashaswi.
yashaswi.@pixperk·
i am suddenly making a go + react webapp
English
33
3
148
10K
Eliza
Eliza@eliza_doesstuff·
@trashh_dev Standard deviation did ruin some of my A/B tests :/
English
0
0
0
225
Eliza retweetledi
Gudim
Gudim@like_gudim·
Gudim tweet media
ZXX
0
10
145
5.4K
Eliza
Eliza@eliza_doesstuff·
@ThePrimeagen It’s better than me. I sometimes leave it open 1/2 day. AI will make me 10x faster!
English
0
0
0
141
Eliza retweetledi
Keith Siau
Keith Siau@drkeithsiau·
The scariest thing in academia 😅
Keith Siau tweet media
English
47
338
2.3K
123.9K
Eliza retweetledi
Andy Boenau
Andy Boenau@Boenau·
"Humans aren’t very efficient movers—until you put us on a bicycle, when we become some of the most energy-efficient land travelers in the animal kingdom."
Andy Boenau tweet media
English
701
2.6K
40.1K
8.2M