Roger Bohn (@RogerBohn) profile picture

Prof. Emeritus @UC San Diego. Also Clinical Excellence Research Center, Stanford. Tech knowledge from craft to science. Currently: healthcare, flying.

San Diego
Joined April 2009
Photo by Roger Bohn (@RogerBohn)Photo by Roger Bohn (@RogerBohn)Photo by Roger Bohn (@RogerBohn)Photo by Roger Bohn (@RogerBohn)Photo by Roger Bohn (@RogerBohn)Photo by Roger Bohn (@RogerBohn)Photo by Roger Bohn (@RogerBohn)Photo by Roger Bohn (@RogerBohn)Photo by Roger Bohn (@RogerBohn)Photo by Roger Bohn (@RogerBohn)
Roger Bohn retweeted
This is the end. Now no one can cross this level !
7
44
10
202
11,905
It's over, folks. Pack it in.
An M.I.T. study found that 95% of companies that had invested in A.I. tools were seeing zero return. It jibes with the emerging idea that generative A.I., “in its current incarnation, simply isn’t all it’s been cracked up to be,” @JohnCassidy writes. nyer.cm/FUZwzw8
740
841
74
16,310
3,190,457
My overall view at this point (slightly more positive than @LuizaJarovsky below) is that generative AI chatbots (distinct from domain-specific tools like AlphaFold) will on net increase productivity to a moderate degree in many domains, with large benefits in a few areas. (I also agree with all of Luiza’s cautions round the costs in making this work.) But those productivity benefits will also come with strong costs to society (misinformation undermining democracy, deep fakes for scams and politics, disruption of high school and college education, delusions that bring some to mental illness, nonconsual deepfake porn, increased cybercrime, greater surveillance, negative environmental impact, etc, to say nothing of an increasingly oligarchical society) Overall, I am not convinced that GenAI will be of significant net benefit to society. I would welcome @emollick’s latest take on this.
🚨 Unpopular opinion: generative AI’s overall productivity gain is probably close to zero. Companies that want to justify their "AI-first" strategy often conveniently ignore the time, effort, and costs involved in making sure AI-assisted work meets minimal standards. They can ignore it, but the additional work will not disappear. 2026 will likely be a year of reality shocks and budget adjustments. - 👉 NEVER MISS my analyses and curations on AI: sign up for my newsletter below (67,500+ subscribers).
170
44
1
182
28,667
X (and other social media sites) make our 1990s optimism about the Information Age seem silly. Even with all of the world's information a click away (& a free AI that can help explain that information in a personalized way), half-mangled anecdotes with no source win every time.
37
55
12
608
58,767
BREAKING! @FrancescaGino's ridiculous defamation lawsuit against @DataColada from last year has been dismissed by the court! This is an important victory.. if this had been allowed to move forward it would have had a chilling effect on whistleblowers! storage.courtlistener.com/re…
12
213
38
1,133
177,314
Yes, college students have lost their ability to read. I have taught lit for 24 years; the threshold started to decline in the late aughts and nosedived during Covid. A thread with observations + how I get my students to read ALL (or at least most) of the reading I assign: 🧵
550
4,679
937
61,750
8,480,675
The choice is simple: 1. A decent but frail old man, surrounded by a good, experienced team. 2. A bat-shit crazy old man, surrounded by a bunch of sycophant hooligans.
6,697
14,804
992
60,429
1,606,172
Underrated info hazard from AI: when employees learn that they can automate their own work & no one cares. I think a lot about this study, when managers doing processes design learned how screwed up their firm was, they gave up trying for leadership roles in existential despair.
25
162
11
1,229
198,449
‘Rochester's three ineffectual investigations should serve as a strong warning to other institutions and departments. Too often, the purpose of such things seems to be to make the problem go away rather than to find out what is actually happening.’
3
4
45
11,105
🧵 Learning intermediate undergrad macroeconomics with @ojblanchard1 and Johnson from a heterodox/HET/methodology angle. That's the text @WilliSemmler first taught me macro with at @NSSRNews. Now that I'm all done with grad school, here's a reading list to accompany the book.
5
32
5
108
42,258
Using the right word for a concept is a cognitive shortcut. Lacking appropriate vocabulary doesn't prevent you from thinking about complex topics, per se, but it increases your cognitive overhead. It's like running with lead shoes. Precise words are tools for thought.
28
75
10
739
68,447
Roger Bohn retweeted
Table of the various 1-year adverse outcomes for both full cohort and those hospitalized along with all-cause mortality graph
2
101
8
392
44,808
An example making clear the distinction between language understanding and likelihood-based text completion.
5
60
8
385
133,094
Roger Bohn retweeted
On the highway towards Human-Level AI, Large Language Model is an off-ramp.
245
283
166
2,700
1,559,531
WHO congratulates #Uganda for ending #Ebola outbreak in less than 4 months. Thanks donors & partners for mobilizing resources & vaccine developers to make available candidate vaccines in record time .
1
7
1
36
1,471
Another use for ChatGPT: correcting OCR errors in historical texts. The quality is quite good
17
132
28
970
203,166
The news about life expectancy in America is devastating. Covid is taking a huge toll, but so are other causes like drug overdoses.
Now at a 25 year low, life expectancy in the US continues to fall. Covid persists as the 3rd leading cause of death and increasing ~20% over 2020 wsj.com/articles/u-s-life-ex… @littlewern @jon_kamp
8
55
2
155
64,292
In order to understand why ChatGPT can't replace Google Search, it's useful to understand the early days of web search and the role that PageRank played. 1/n
48
316
57
1,913
736,456
Please stop focusing on learning from ultra-successful founders. Survivorship bias is acute among these folks & the outliers often are those making big risky bets that happen to pay off. And even if that isn’t the case, we usually learn the wrong lessons. oneusefulthing.substack.com/…
5
60
13
331
83,576