Ola Wale (INFJ) (@clockpenalty) profile picture

I co-host a podcast and write at @TheNaijaNerds. I also write comics for @comicrepublicng

London
Joined January 2021
Representatives from Comic Republic paid a visit to the Director General National Orientation Agency, Mallam Lanre Issa-Onilu at the agency’s headquarters in Abuja to discuss potential collaboration opportunities. The representatives from Comic Republic expressed their interest in working with NOA with the objective of fostering and instilling patriotism within the society, promoting Nigerian heritage, unity, social norms, educating and encouraging Nigerians to uphold national values, and championing a global and national strategy campaign on Nigeria's National Value Charter. They also expressed their interest in serving as ambassadors of the agency to educate young Nigerians on the National Value Charter. The DG acknowledged the alignment of Comic Republic's objectives with NOA's mission, emphasizing the importance of institutionalizing values rather than mere sloganeering. He highlighted the need to create relatable indigenous characters to counter the dominance of foreign cartoons. #NOA #NOANigeria #ComicRepublic #partnerships
11
44
3,957
"A Taxonomy of AI Panic Facilitators" A visualization of leading AI Doomers (X-risk open letters, media interviews & OpEds). Some AI experts enable them, while others oppose them. The gender dynamics are fucked up. It says a lot about the panic itself. Your thoughts?
111
166
61
669
427,602
Ola Wale (INFJ) retweeted
Pro tip: if a machine learned model scores 100% on something, the metric is wrong, or the evaluation dataset is not robust, or the evaluation dataset is in the training data, or a combination of all of the above. At least in this point in time.
This quoted post is unavailable.
45
154
17
1,069
219,824
Absolutely brilliant!
This Post is from a suspended account. Learn more
2
1
57
Ok guys, please listen. LLMs have no memory, no recall of past events, no mutable internal state. They are complicated functions that map input strings to output strings. They are incredible nonetheless, no need to overcomplicate things. And no, humans are not "maybe also that".
92
113
18
825
175,239
I think it's absolutely wild that I can literally run a fully functional AI language model at full speed on my PC with no internet connection and 0 GPU utilisation. Insane that the same model can literally run on a cellphone. I have to say I didn't expect this pace of progress.
1
33
Ola Wale (INFJ) retweeted
LLaMA is a new *open-source*, high-performance large language model from Meta AI - FAIR. Meta is committed to open research and releases all the models the research community under a GPL v3 license. - Paper: research.facebook.com/public… - Github: github.com/facebookresearch/…
89
397
112
2,275
915,433
Wale Awelenje of @comicrepublicng is another shining light on the list. He made magic with his work on Vanguard Outreach. You're missing out if you haven't read this
1
3
5
170
Ola Wale (INFJ) retweeted
LLMs are still making sh*t up. That's fine if you use them as writing assistants. Not good as question answerers, search engines, etc. RLHF merely mitigates the most frequent mistakes without actually fixing the problem.
the success of chatgpt has lead to investors thinking RLHF is magic (to some extent it is), but boy they are going to be disappointed when their portfolios realize its limitations
48
191
43
1,070
435,201
Ola Wale (INFJ) retweeted
I do not want to hear one more successful person in an elite profession saying "well I'm innumerate and it never harmed meeeee" when there are millions of people out there getting mugged off because they can't do simple financial calculations
42
46
5
607
47,766
Ola Wale (INFJ) retweeted
Friendly reminder that the "drone army of data-entering robots" (or "people who use maths in their jobs", as I prefer to call them) are the reason you have abundant food, medical treatments that work, warm secure buildings, and the technology that screens your films
137
276
67
2,921
435,136
If Microsoft actually succeeds in driving adoption of Bing using AI to steal material share from Google then there’s no question Satya becomes the greatest CEO of all time
37
89
13
1,965
349,580
Nobody knows how LLMs or Diffusion models work. We only know how to find them. We call the finding process "training" because hubris. When asked how they work, we overcomplicate describing the "training" so they nod, confused, trusting we know what we are doing. We don't.
36
Ola Wale (INFJ) retweeted
Current human quests summed up: - create and enslave a God - create and control a Sun - cheat Death
69
560
143
4,497
895,143