Shuki cohen (@CohenShuki) profile picture

A passionate Data Scientist, thrilled by the power of Machine & Deep Leaning to reveal unintuitive insights on human nature

Joined July 2020
Photo by Shuki cohen (@CohenShuki)Photo by Shuki cohen (@CohenShuki)Photo by Shuki cohen (@CohenShuki)Photo by Shuki cohen (@CohenShuki)Photo by Shuki cohen (@CohenShuki)Photo by Shuki cohen (@CohenShuki)Photo by Shuki cohen (@CohenShuki)Photo by Shuki cohen (@CohenShuki)Photo by Shuki cohen (@CohenShuki)
Shuki cohen retweeted
Are you coming to @DeepLearningAI AI's AI Dev 26 x SF next week? Our Chief Product & Strategy Officer, Or Dagan, will be giving a talk diving into what our product and R&D teams are doing to automate finding the optimal balance between cost, accuracy, and latency for every agent that we build. Check out the teaser below and head to Stage 2 at 4:15 pm on 4/29!
1
1
7
84,897
Shuki cohen retweeted
1/5 Tricking Parsers: Improving RTL Parsing by Faking English: PDF document parsing is an important task for RAG. However, there are gaps in the capabilities. Especially when it comes to underrepresented languages (Results in pink show our improvements to the SOTA). @AI21Labs ๐Ÿงต
1
3
10
821
ืื™ืš ื ืจืื” ืžืจื•ืฅ ื”ื—ื™ืžื•ืฉ ื”ืขื•ืœืžื™ ืœื‘ื™ื ื” ืžืœืื›ื•ืชื™ืช, ืžื ืงื•ื“ืช ืžื‘ื˜ื” ืฉืœ ื”ื—ื‘ืจื” ื”ื™ืฉืจืืœื™ืช ื”ื™ื—ื™ื“ื” ื‘ืžืจื•ืฅ? @AI21Labs ๐Ÿ‡ฎ๐Ÿ‡ฑ ื”ืฆื˜ืจืคื• ืืœื™ื™ ืœืฉืจืฉื•ืจ ืฉื™ืชืŸ ืœื›ื ืคืจืกืคืงื˜ื™ื‘ื” ื˜ื™ืคื” ืื—ืจืช, ื•ื”ืจื‘ื” ื™ื•ืชืจ ืžืงื•ืžื™ืช ๐Ÿงต...
2
1
5
377
ืœืžื•ืœ ื–ื”, ื”ืฆื•ื•ืช ืฉืœื ื• ืฆืจื™ืš ืœืชืช ืžื•ืฆืจื™ื ื•ืคืชืจื•ื ื•ืช ืชื—ืจื•ืชื™ื™ื, ื•ื–ื” ืจื‘ื•ืชื™ื™, ืœื ืคืฉื•ื˜ ื‘ื›ืœืœ. ื“ืžื™ื™ื ื• ืฉืืชื ืžื—ื›ื™ื ืœื”ื™ื›ื ืก ืœืžืฉื—ืง ื›ื“ื•ืจื’ืœ ื•ืฆืจื™ื›ื™ื ืœื‘ื—ื•ืจ ื‘ื™ื ื™ื›ื ืœื‘ื™ืŸ ืžืกื™ ืื• ืจื•ื ืœืื“ื•, ืืช ืžื™ ื™ื‘ื—ืจื• ืœืงื‘ื•ืฆื”? ืื– ื›ื–ื”. ืื‘ืœ ืื ื—ื ื• ืžืชื ื”ื’ื™ื ื›ืžื• ื™ืฉืจืืœื™ื ื˜ื•ื‘ื™ื - ื”ืคื•ื–ื™ืฆื™ื” ื”ื–ื• ืจืง ืžื“ืจื‘ื ืช ืื•ืชื ื• ๐Ÿ˜Ž
1
88
ื›ืฉืžืชืกื›ืœื™ื ืฉื•ื‘ ืขืœ ื”ื™ืฆืข ื”ืžื•ื“ืœื™ื ืฉืœ Bedrock ืฉืœ AWS, ืžืจื’ื™ืฉื™ื ื’ืื•ื•ื”. ืื™ืฉื™ืช, ืื ื™ ื”ื‘ื ืชื™ ื‘ืฆื•ืจื” ื’ื•ืœืžื™ืช ื›ืžื” ื”ื—ื‘ืจื” ืฉื‘ื” ืื ื™ ืขื•ื‘ื“ ื”ื™ื ื—ืœื•ืฆื” ื™ืฉืจืืœื™ืช, ืคื•ืจืฆืช ื“ืจืš ืขื ื“ื’ืœ ื›ื—ื•ืœ ืœื‘ืŸ ืชืงื•ืข ืขืœ ื”ื’ื‘ ืฉืœื”. ื•ืื™ื–ื” ื”ื™ืฉื’ ืขื ืง ื”ื•ื ืœื”ื’ื™ืข ืœื‘ืžื” ื”ืจืืฉื™ืช ืฉืœ AWS, ืขื ืงื™ืช ื”ื˜ื›ื ื•ืœื•ื’ื™ื” ื”ืขื•ืœืžื™ืชโ€ฆ
83
Shuki cohen retweeted
Meet Maestro: @AI21Labs new AI Planning and Orchestration system. Unlike LLM-based agents, #Maestro ensures reliable outcomes while optimizing latency and cost. This is AI that you can actually trust. ๐Ÿ”— Sign up now! ai21.com/maestro #AI21Maestro #AccuracyAtScale #AIOrchestrator #AIPlanning
1
25
9
96
170,967
Shuki cohen retweeted
Great 6 minute in depth of Apple Intelligence. Quantized models running with finetuned adapters on device, with a handoff to private cloud, including a security overview. Very well executed, OpenAI nowhere to be found ๐Ÿ‘€
5
17
8
111
29,842
New York skies, last night ๐Ÿ’”๐ŸŽ—๏ธ #BringThemAllHome #Bringthemback
33
266
21
1,263
33,445
Shuki cohen retweeted
We just released Jamba-Instruct! Built from our groundbreaking SSM-Transformer Jamba architecture, Jamba-Instruct brings the same technological innovation to the enterprise via an aligned model. With leading quality benchmarks, a 256K context window, and the most competitive pricing in its size class, youโ€™re getting the most value for your money. Read more in our blog: ai21.com/blog/announcing-jamโ€ฆ (1/4)
10
54
9
271
42,078
Shuki cohen retweeted
He knows what heโ€™s talking about
7 years ago, the UAEโ€™s Foreign Minister issued a warning to the West. His words now sound prophetic.
11,811
71,374
3,175
343,883
60,569,593
ืื•ื™ ื–ื” ื ื”ื“ืจ - ื”ืคื™ื™ืกืจืก๐Ÿ€ ื”ืฉืชืžืฉื• ื‘ืคื™ืœื˜ืจ ื‘ื›ื™ ืฉืœ ืกื ืืคืฆืื˜ ืขืœ ืงื”ืœ ื”ืื•ื”ื“ื™ื ื”ื™ืจื™ื‘ื™ื (ืœื™ื™ืงืจืก) ืชื•ืดื› ื”ืžืฉื—ืง...
4
5
4
723
53,871
Shuki cohen retweeted
Introducing Jamba, our groundbreaking SSM-Transformer open model! As the first production-grade model based on Mamba architecture, Jamba achieves an unprecedented 3X throughput and fits 140K context on a single GPU. ๐Ÿฅ‚Meet Jamba ai21.com/jamba ๐Ÿ”จBuild on @huggingface
35
237
76
1,091
333,101
GRAPHIC: EVIL: Since a NY Post journalist already released this, it is something so evil that it's hard to write. They baked a baby in the oven. The evil monsters baked a baby in the oven.
820
3,817
557
8,638
2,713,160
Indeed. An inability to distinguish between terrorists trying to kill as many civilians as possible and a military operation against a terrorist force operating among civilians is a capitulation to barbarism. Mourn every innocent death, but do not confuse cause and effect.
NOTE TO NEW PALESTINE TWITTER EXPERTS: Hamas launching a 1,500 man terrorist force and literally Invaded a neighboring nationโ€™s borders, then individually executed 1,400 civilians irregardless of race, religion or nationality in cold blood This is not self defense or even resistance. It was literally an attempted genocide. No one they encountered was spared. #ThatIsAll
417
2,582
335
9,914
1,872,509