Deepak George

668 posts

Deepak George

Deepak George

@deepak_george

AI Leadership, AI/ML aficionado, IIM Bangalore, College of Engineering Trivandrum. Tweets/Opinions are my own views

Netherlands เข้าร่วม Haziran 2009
1.5K กำลังติดตาม268 ผู้ติดตาม
Ubisoft UK
Ubisoft UK@Ubisoft_UK·
Giveaway time! This Holiday, we are giving away a Star Wars Outlaws Switch 2 digital code. How to enter the Giveaway: Like this post Comment Star Wars Outlaws Giveaway ends on 24/12/2025 12:00PM GMT and is only open to residents of the UK.
Ubisoft UK tweet media
English
1K
168
1.4K
72.1K
Deepak George
Deepak George@deepak_george·
@madebygoogle @sundarpichai Pixel have swollen battery & it pushed the screen out. Without a phone for > 1 week. Delivery due date for replacement phone was 6 Sep but its NOT even shipped. Expected Google to treat customers way better. Next time going to get iphone RMA88223280
English
1
0
1
138
Made by Google
Made by Google@madebygoogle·
@deepak_george Hi Deepak, we truly apologize for the inconvenience. We'd like to assist you on this over DM. Please click on the link below to send us a direct message. Rest assured, we'll do our best to help you further. Appreciate your understanding. ^Lisa twitter.com/messages/compo…
English
1
0
0
42
Deepak George รีทวีตแล้ว
Damien Henry
Damien Henry@dh7net·
Someone asked me recently why the Paris AI ecosystem is so 🔥 these days. On the surface, it looks like Paris became a major #AI center overnight, but it didn't. It took time and didn't appear by accident. Here is a story that started more than 10 years ago. 👇
English
17
104
575
241.2K
Deepak George รีทวีตแล้ว
the tiny corp
the tiny corp@__tinygrad__·
We are not AMD's QA team, and we have no relationship with them. I saw some stuff last year that gave me hope, but this platform has been out for 14 months now, and there's still serious issues. It upsets me that the MES isn't open source. While more stuff is open source than NVIDIA, if there's blobs we do not own the hardware, and I don't feel great about investing time into this. The compiler bug today is icing on the cake. At first I thought it was the `launch_bounds` feature, but it looks like it can be triggered without that. Not being able to trust a compiler undermines so much trust in the entire platform. It would set us back, but maybe we should switch to either 3090s or @intel GPUs. Either way, we aren't shipping the tinybox (or ordering bulk 7900XTX) until this is figured out.
English
9
8
226
22K
Deepak George
Deepak George@deepak_george·
@SinnerOnPC Is reverse split to push stock above $1 allowed by nasdaq when they have warning from nasdaq already?
English
1
0
0
43
Sinister964
Sinister964@SinnerOnPC·
$ME should go private or delist. 🤷‍♂️ I think we all know it's going to reverse split at this point. I'm glad I didn't trade it.
English
1
0
1
188
Deepak George รีทวีตแล้ว
Made by Google
Made by Google@madebygoogle·
iPhone® leaves a big case to fill, but Pixel is ready to live up to the legend. Watch the full #BestPhonesForever series on our YouTube channel: goo.gle/45h5pIH
English
329
635
3.7K
731.1K
Deepak George
Deepak George@deepak_george·
@soumithchintala @FedorShabashev Agree that Nvidia's cuda is not required for inference and cane be done on CPU. But what about fine tuning LLM models? These still heavily depends on CUDA since alternatives like ROCM and OneAPI are not officially upstreamed to pytorch and tensorflow yet.
English
1
0
1
701
Soumith Chintala
Soumith Chintala@soumithchintala·
@FedorShabashev it doesn't matter if AMD is very far for CUDA adoption -- that doesn't matter for inference on a narrow set of workloads. You can even hand-write the inference code in assembly if the workloads are homogeneous and high-volume.
English
6
1
77
12.8K
Soumith Chintala
Soumith Chintala@soumithchintala·
I'm fairly puzzled by $NVDA skyrocketing. GenAI inference and fine-tuning will significantly outweigh GenAI training in overall compute. When it comes to inference and fine-tuning, NVIDIA's advantage in software won't hold much significance. They will inevitably have to face competition from AMD and custom silicon in terms of performance per dollar. As a result, their high-profit advantage will inevitably diminish. The software work that enables this moat-shedding is @MosaicML Foundry, @huggingface Optimum, @OpenAI Triton, PyTorch 2.0, @Microsoft ONNXRT (and maybe eventually @Modular_AI). These stacks abstract hardware out without change in functionality or reliability. I believe NVIDIA will enjoy their high-margin advantage for another 12 to 18 months before they are forced to make adjustments.
English
73
132
995
417.7K
Deepak George รีทวีตแล้ว
Kerala Blasters FC
Kerala Blasters FC@KeralaBlasters·
Still trying to figure out 𝗪𝗛𝗘𝗡 the referee whistled or marked the distance of the wall. #BFCKBFC #ഒന്നായിപോരാടാം #KBFC #KeralaBlasters
English
886
1.3K
9.9K
763.2K
Deepak George
Deepak George@deepak_george·
@heymaxkoh Any other book you recommend for valuating/investing in tech companies?
English
0
0
0
0
Deepak George
Deepak George@deepak_george·
@GergelyOrosz Why won't valuations of these private pre IPO companies also fall 70-80% like public competitors?These pre ipo companies would also be forced to cut valuations by 70% if they wanna raise money from investors in the current market correct?
English
0
0
0
0
Gergely Orosz
Gergely Orosz@GergelyOrosz·
Late-stage, VC funded companies which have not cut their valuations by ~70-80% are now the worst of both worlds: - Illiquid equity (public companies: equity is liquid) - Little to no upside (early-stage startups: far more upside)
English
4
7
192
0
Gergely Orosz
Gergely Orosz@GergelyOrosz·
A massive change of the past ~6 months is how the “conventional wisdom” of the surest way to make big money working in tech is to join a late-stage, VC-funded company, pre-IPO. Well, this no longer applies. In fact; this route is now one of the worst *financial* choices.
English
15
31
496
0
pauluaq
pauluaq@powlwoq·
Tesla shows end of September delivery date on Model Y RWD in Germany, ordered end of August 👀 @SawyerMerritt @wholemars
pauluaq tweet media
English
16
11
162
0