Olalekan Taofeek (@harphies) header
Olalekan Taofeek (@harphies) profile picture

AI/ML Infrastructure Engineer Building Infrastructure for the next revolution of technology More active on LinkedIn to see more contents from me.

London, UK
Joined August 2017
Photo by Olalekan Taofeek (@harphies)Photo by Olalekan Taofeek (@harphies)Photo by Olalekan Taofeek (@harphies)Photo by Olalekan Taofeek (@harphies)Photo by Olalekan Taofeek (@harphies)Photo by Olalekan Taofeek (@harphies)Photo by Olalekan Taofeek (@harphies)Photo by Olalekan Taofeek (@harphies)Photo by Olalekan Taofeek (@harphies)Photo by Olalekan Taofeek (@harphies)
Your health deserves to be understood, not guessed at. So TreatwiseAI shows you its thinking. Every step, connecting your heart rate to your stress, your stress to your sleep. Because the dots you connect early are the ones that never become something serious. 🧵
🤖 Made with AI
1
1
2
62
Olalekan Taofeek retweeted
8
593
27
4,170
174,493
Olalekan Taofeek retweeted
Furthermore, this whole thing is completely false — none of this is happening. The old adage “don’t believe everything you read” is even more true today than it ever has been. Now lies can get ALL the way around the world before the truth can get its pants on. So be careful out there folks and don’t be gullible. Will be interesting to see if all the outlets that “covered” and re-reported on this issue a correction when it comes and goes and doesn’t happen.
This is not credible. Unless you are buying each of your guests a house, you can’t spend this much money.
4,706
2,616
1,179
40,765
10,283,245
Olalekan Taofeek retweeted
How do you test critical features when you lack a dedicated dev environment? Or when you need to validate changes against production configurations? check out my article on implementing preview deployments with Argo CD pull-request generator. 1/2
2
5
11
599
Many talented people don't live up to their potential because they are consumed by unhealthy relationships If you keep choosing people who only make your life harder, you will inevitably feel exhausted You will unlock your potential once you understand who you need to be around
46
288
15
2,114
69,083
Olalekan Taofeek retweeted
“Talented, dedicated teams willing to take risks to make something happen… that’s the scarce commodity - not money.” ~ Elon Musk
23
395
42
2,770
190,040
Olalekan Taofeek retweeted
54
5,196
137
33,399
1,903,754
Olalekan Taofeek retweeted
33
2,578
125
15,530
923,615
103
7,464
185
37,535
3,752,370
Llama 3.1 and open source are now competitive with frontier proprietary models. However, when considering the relative price-positioning of Llama 3.1, while 70B is achieving a clear leadership position, 405B is being offered at a higher price by some providers relative to proprietary models. We explore why this is the case below 👇 With an MMLU score of 89%, Llama 3 is in-line with the leading proprietary models including GPT-4o (89%), Claude 3.5 Sonnet (89%) and Gemini 1.5 Pro (86%). Furthermore, the Llama 3.1 series also supports long context windows (128k) which was previously mainly confined to proprietary models (with exceptions such as Qwen 2). Overall, this means frontier model capabilities are accessible within open-source and across many providers, rather than being reliant on the proprietary offerings of AI labs. However, when comparing Llama 3 405B in price with the price that GPT-4o and Claude 3.5 Sonnet are offered at, the median provider is offering Llama 3.1 405B at a higher price. In prior tweets we have detailed why we expect this is the case. Llama 3 405B at FP16 cannot ‘fit’ on a single DGX system with 8xH100s, it requires 2, increasing the cost dramatically. We expect over time, providers will serve with other hardware and model configurations to manage the cost and bring their 405B offering into a price competitive position. Configurations we expect include serving the model on AMD’s MI300X chips and serving a quantized version of the model at lower precision (FP8) to fit the model on a single server at the cost of a mild quality degradation. Congratulations to @GroqInc , @FireworksAI_HQ , @togethercompute, @replicate , @awscloud , @databricks on the fast launch of @AIatMeta's Llama 3.1 models! We have commenced our speed benchmarking and running our independent quality evaluation suite across providers serving the models and will be sharing results shortly - stay tuned. We will also be following up with a tweet looking at individual provider pricing in-depth. See below for a link to our analysis of providers serving Llama 3.1 👇
2
9
83
5,178
Olalekan Taofeek retweeted
Replying to @DrInTech22
Virtualisation and all that container stuffs is just to manage hardware resources better. Thanks to technology like kubernetes and other virtualisation technologies in the data centre that abstracts most of those complexity to the software layer. The world is really trying to solve hardware problems with software but that has a limit and we can all see that with AI and the high demand for compute power almost everyday.
1
3
7
404
Olalekan Taofeek retweeted
Learn more about DripX. By it notification, it keeps you ever close to your patient. To request a demo , TrepLabs.co
4
2
832
Paul Graham explains why you shouldn’t try to be a visionary “Empirically, the way to do really big things seems to be to start with small things and grow them bigger. Want to dominate microcomputer software for decades? Start by writing a basic interpreter for a machine with a couple thousand users. Want to make the universal website and a giant vacuum for people’s time? Start by building a website where Harvard undergrads can stalk one another.” Paul Graham continues: “Neither Bill Gates nor Mark Zuckerberg knew how big their companies were going to get. All they knew was that they were onto something… Maybe it’s a bad idea to have really big ambitions initially, because the bigger your ambitions, the longer they’re going to take to realize and the long you’re projecting into the future, the more likely you’re going to be wrong.” PG suggests starting with something small that works instead. “I think the best way to do these big ideas is not to try and identify a precise point in the future and say, How do I get from here to there? Like the popular image of a visionary. I think a better model is Columbus who thought there was something to the West—I’ll sail westward. Start with something that works, that you know works, that’s small, and then when the opportunity comes to move, move westward. The popular image of a visionary is someone with a very precise view of the future, but empirically it’s probably better to have a blurry one.”
15
164
18
790
92,743
Olalekan Taofeek retweeted
It's a new week 💕🤩. Want to learn something new for the week? Check out our post.
4
8
484
You can waste years achieving nothing despite having everything you needed: time, talent, capital, connections, material comfort, because what you actually needed was the hunger that comes from not having what you need.
39
437
6
2,654
126,231
Olalekan Taofeek retweeted
More about DripX
5
7
439
Olalekan Taofeek retweeted
“I don’t ever give up. I’d have to be dead or completely incapacitated.” ~ Elon Musk interview on this day in 2012.
30
450
38
2,730
247,559
Olalekan Taofeek retweeted
AI is revolutionizing ! From diagnostic tools to treatment. TrepLabs is revolutionary. Welcome to a new week , embrace AI based health devices.
3
7
302
Olalekan Taofeek retweeted
Know this : A Blood Transfusion might not be lifesaving if it's done wrongly.
2
6
277
Olalekan Taofeek retweeted
Self-Discover is a fascinating new algorithm from researchers at Google DeepMind and USC that searches for "atomic reasoning modules" One of the quickest ways to improve your LLM programs is to add Chain-of-Thought, "Let's think step by step ...", another primitive like this could be "Let's think out of the box" or "Think symbolically to calculate an answer..." Self-Discover searches for these thinking primitives that work best for YOUR LLM programs!! 🚀 I am SUPER excited to publish our 90th Weaviate podcast with @cdossman, a serial entrepreneur who has implemented Self-Discover in DSPy!! 🤯🔥 I think this is such a fascinating example of the potential of DSPy and it was so much fun to catch up with Chris who joined the Weaviate Podcast alongside @bobvanluijt and @marcobianco on the 30th podcast, now back for the 90th!! 🎙️🚀 Link: youtube.com/watch?v=iC64q1gF…
8
37
5
159
23,102