Adam Sadovsky
@asadovsky
Followers
770
Following
328
Media
0
Statuses
89
Distinguished Software Engineer / Senior Director, Gemini
Joined December 2007
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Butler
• 705058 Tweets
الهلال
• 637702 Tweets
#KadınaVeÇocuğaDokunma
• 287645 Tweets
#TurkishWomenNeedHelp
• 165592 Tweets
#WWEBadBlood
• 145536 Tweets
Drew
• 95491 Tweets
Milton
• 71606 Tweets
Carvajal
• 64291 Tweets
Phillies
• 63307 Tweets
Yankees
• 48306 Tweets
Vini
• 29285 Tweets
Vandy
• 26607 Tweets
جيسوس
• 25727 Tweets
Bama
• 24808 Tweets
#precure
• 24465 Tweets
Wheeler
• 23489 Tweets
Iowa
• 23191 Tweets
Dark MAGA
• 22985 Tweets
الشوط الاول
• 22222 Tweets
Handi
• 21460 Tweets
Ohio State
• 18854 Tweets
فهد العتيبي
• 17458 Tweets
Vanderbilt
• 16001 Tweets
الشوط الثاني
• 15807 Tweets
Bayley
• 14748 Tweets
Hell in a Cell
• 13167 Tweets
Ave Maria
• 12997 Tweets
Macron
• 11333 Tweets
Hoffman
• 10064 Tweets
Gemini Advanced is here! And it's not lazy. Let us know what you think!
21
8
82
Now Gemini Pro is coming today in Bard’s biggest update yet (in English in 170 countries) with more advanced reasoning and understanding in the responses. Bard Advanced with Ultra, our most general and capable model for highly complex tasks, is coming early next year.
60
204
2K
3
5
55
@saqbach
Different fine-tuning, and Bard's has access to the internet. We are working hard to improve both versions of the model though, stay tuned!
4
3
44
🧑🍳
Exciting News from Chatbot Arena!
@GoogleDeepMind
's new Gemini 1.5 Pro (Experimental 0801) has been tested in Arena for the past week, gathering over 12K community votes.
For the first time, Google Gemini has claimed the
#1
spot, surpassing GPT-4o/Claude-3.5 with an impressive
84
419
2K
2
0
36
oh hai
🔥Breaking News from Arena
Google's Bard has just made a stunning leap, surpassing GPT-4 to the SECOND SPOT on the leaderboard! Big congrats to
@Google
for the remarkable achievement!
The race is heating up like never before! Super excited to see what's next for Bard + Gemini
153
621
3K
2
1
28
1
0
15
the secret is out
0
0
15
It's a good model! 🧑🍳🧑🍳🧑🍳
Dang, the new Gemini model is quite good
I gave it an excellent, complicated game source book PDF and asked it to roll up a character, referencing many pages. There may be errors, but wow. First time I have seen an AI get so far
Interestingly, it decided to play... a secret AI
15
55
526
1
0
14
@JackK
@pra15meshh
@AarushSelvan
So, this one is a bit funny actually. The new Gemini UI is so clean and simple that the model focused its attention on the first example prompt, which happened to be about rusty plants. Of course, we should do better! But, it's not completely random. 😅
0
0
13
Very nice work. Warren Buffett's famous quote seems apt: “Only when the tide goes out do you learn who has been swimming naked.”
Data contamination is a huge problem for LLM evals right now. At Scale, we created a new test set for GSM8k *from scratch* to measure overfitting and found evidence that some models (most notably Mistral and Phi) do substantially worse on this new test set compared to GSM8k.
36
216
1K
1
0
11
A big step forward. LLMs are much more useful when they're easy to fact-check.
Excited to launch and share the new citations feature ("related content") in Gemini! Last fall, we introduced Double-Check (), which checks Gemini's claims against sources on the web. With citations, we're now auto-running on all fact-seeking queries.
1
6
39
0
0
10
The momentum is palpable!
Two new production Gemini models, >2x higher rate limits, >50% price drop on Gemini 1.5 Pro, filters switched to opt-in, updated Flash 8B experimental model, and more.
It’s a good day to be a developer : )
176
324
2K
0
0
8
@bindureddy
Some other LMSys-ranked models like pplx-70b-online also have internet access, but it's true that many don't. Internet access can both help and hurt; a model must be tuned carefully to make good use of it. We'd love to see other attempts at it!
0
0
7
👀
[Arena] Exciting news from Arena! We're thrilled to welcome Google's Bard, now enhanced with Gemini Pro for advanced summarizing, reasoning, and real-time insights.
Curious how Bard measures up against leading AI models? Join us in the Arena to find out!
7
23
183
0
0
6
Gemini is now available to businesses large and small. Let us know what you think!
We’re launching Gemini Business for
@GoogleWorkspace
. This new add-on gives organizations of all sizes access to Gemini for Workspace, including Help me write in Docs, image generation in Slides and more ↓
160
158
882
0
0
5
👀
For the first time, I am seeing signs of "cleverness" in Bard, which has been upgraded to Gemini Pro (GPT-3.5 levels of reasoning)
I gave it the apple test - ending sentences with the word apple. It still mostly fails, but when asked to visualize it, the results are interesting
11
20
209
0
0
5
@Paul_Moreira
We have some filters that are too aggressive and imprecise. We're working hard to fix it, stay tuned!
1
0
3
@permaximum88
It's been really helpful to have real users "kick the tires" on it, as we did for Gemini Pro back in December. But we'll get it there in due time!
0
0
1
@permaximum88
@JackK
@samc0hen
@mbaeuml
Ironically, the response from your screenshot is completely wrong - check the numbers.
The free version of Gemini also has access to up-to-date information, so this isn't about free vs. paid. We just have some work to do to answer these prompts reliably. We'll get there!
1
0
2
@UltraRareAF
@hu_yifei
Yeah, it's a challenging space. We're thinking about how best to handle this. Stay tuned.
2
0
1
@hevensxj
@Paul_Moreira
I agree, they make my blood boil too. Long story for why it happens currently, but we'll fix it!
@JackK
1
0
2
Feels like half of LLM "examples that will blow your mind" have major mistakes in the output when you look closely.
0
0
2
@hu_yifei
Ugh, sorry. I got a slightly better response when I tried:
At any rate, this is bad. I've shared it with the team. We have an active effort to fix these kinds of failures.
2
0
2
1
0
1
@Jason
@sundarpichai
@Jason
- we took your UI feedback to heart... let us know what you think of our new Gemini UI!
1
0
2
@drjwrae
Agreed. Post-training makes all the difference for making the model useful for end users.
0
0
2
@OfficialLoganK
@Google
Congratulations and welcome to Google - thrilled that you're joining and can't wait to see what you build!
0
0
1
@dharmesh
Try out the Bard+Workspace integration! Bard can search your Gmail and answer your questions. Details:
0
0
0
@DimitrisPapail
@altryne
@lmsysorg
@Google
It's the best public leaderboard we know of, but please let us know if there's another we should look at.
0
0
1
@Klotzkette
@mbaeuml
@ChatGPTapp
@MSFTCopilot
Improving accuracy is an ongoing major focus. If you're open to it, please DM me examples of the problems you're seeing.
0
0
1
@JackK
@BounceNStretch
@altryne
@lmsysorg
@Google
+1, this is a known issue and we're working on it!
0
0
1
@permaximum88
@JackK
@samc0hen
@mbaeuml
Thanks for the feedback! We definitely want to nail these prompts.
0
0
1