Roman Pshichenko
@romechenko
Followers
1,295
Following
60
Media
512
Statuses
4,135
exploring AI and Tech. Software Engineer @sofi
Sacramento, CA
Joined January 2014
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Milton
• 1020493 Tweets
Florida
• 922469 Tweets
#ファミマのスイーツパン誕生
• 131633 Tweets
Chiefs
• 115320 Tweets
Saints
• 85232 Tweets
サンデー
• 78048 Tweets
#HaryanaElectionResult
• 59705 Tweets
Yankees
• 41605 Tweets
Carr
• 36337 Tweets
Real Facts
• 31445 Tweets
जम्मू कश्मीर
• 31021 Tweets
Mahomes
• 30595 Tweets
Asaramji Bapu Case
• 27705 Tweets
Glock
• 26003 Tweets
Olave
• 14375 Tweets
@diegojimenez
Here’s one with some fixes.
1) look over the shortcut to understand what it’s doing
2) add this shortcut
3) create a personal automation to run this shortcut when airplane mode it turned on
Fixes: cancel button exploit
18
40
585
Lovin' creating images in a style I like! Used iOS to crop the background out. Check out these designs - perfect for T-shirts!
#creative
#Tshirts
#iOS
#midjourney
3
17
258
@mattzollerseitz
There is hardly any occasion where I can watch a movie at home with the volume up high. I need to be able to hear my kids and/or spouse. Maybe if it’s late enough, I can watch something with headphones
6
0
174
@anammostarac
All the OpenAI interactions are treated as external tools and need your approval before any information is shared. Other models will be options eventually
13
3
153
@SadlyItsBradley
Like he explained, reviews aren’t for the company/brand, reviews are for the people. So if there’s some misunderstanding then he’s addressing his audience directly
0
0
131
@MatthewBerman
@_akhaliq
Nope, you can actually “reshoot” as many times as you want and with as much control as you want. It’s basically a 3D scene and you get to control the camera.
4
4
73
@sundarpichai
Sundar, is the voice and music in the video also generated by the model? The picture Jeff posted seems to imply the output can only be text and images
We designed Gemini to be multimodal from the start, rather than starting with a purely text model and then grafting on vision and audio encoders after the fact. This means you can naturally interleave modalities: say a few things, add an image, add some text, maybe a short
8
72
600
5
1
68
@evidence444
@levelsio
@Piero_Go_
@nirsd
Because interior designers can own the final process and results. Your goal is not to get a picture. Your goal is to transform your space executed under constraints. Execution and optimization is on designer.
1
1
69
@karpathy
That would explain why zooming in on something just reveals more details. Were the details there before you looked?
15
1
66
@diegojimenez
It works by saving a time stamp of last time you entered secret code. If it was less than 15 seconds ago, it lets you set airplane mode without running the shortcut again.
1
0
58
@jacobin
Even if the comments about jacobin paying less that Walmart are true, this is an ad hominem attack—it doesn’t address the claim against Walmart
78
0
50
@nimbleglide
@Europarl_EN
I hardly ever miss the 3.5mm
Would you really want to use it?
Were there good reasons to get rid of it?
10
0
48
@MichaelEMcNeil
@fermatslibrary
Wouldn’t it be (100000 meteors / hour) * (1 hour / 3600 second) ≈ 27.78 meteors / second
27 meteors showing up every second leaves a lot of room for overlap.
4
0
48
GPT will not only pretend to be a computer for you but will also let you pretend that there’s a python based tic-tac-toe game on the machine that you can play.
2
7
38
@gbstar999
@Noahpinion
@BlueCheese608
@scienceisstrat1
Yeah, I did not know! What else are they doing? This is quite a result
3
1
32
@vincentriemer
I got it to play text-based tic-tax-toe with me.
GPT will not only pretend to be a computer for you but will also let you pretend that there’s a python based tic-tac-toe game on the machine that you can play.
2
7
38
1
2
36
I'm exploring NPC's with memories and motives. Not just memories of their life but memories of interacting with you before. These are dynamic and engaging characters. My approach is to use LLM's like GPT-3 to build a "Thought Process". See proposal. 1/3
2
7
33
@DrJimFan
How hard would it be to train a couple hands on a rail to cook food if mounted in my kitchen?
3
2
31
@dr_cintas
I tried to take a shortcut by having it generate a picture directly based on my image. I got in trouble 😂
1
0
31
0
0
29
@ash_lmb
Here’s a shortcut I made that I’ve been liking for turning <40 minute voice memos into Apple notes. It uses openai whisper and gpt 3.5
@michael_nielsen
@gdb
Here’s what I’ve been using. A shortcut that uses whisper to transcribe the audio then GPT-3.5-turbo to make a summary, title and action items. You can run it directly from voice memos. Handles 30 minute clips in a minute or two.
2
2
25
0
2
29
@chrisprucha
I’ve been banning accounts left and right, muting some. Almost all have hundreds of thousands of followers but just repost memes or even straight up death. I’m starting to see a bit of light at the end of the tunnel but so much to clean 🧼 I used to love For You
5
0
27
@sonjathegrey
@MatthewBerman
@_akhaliq
Generally nerfs are trained by providing lots of images of the scene from many angles. Once the nerf is trained, it’s as good as a 3D scene. It’s better than a 3D scene in a lot of ways since it captures really rich lighting information.
2
1
28
@elonmusk
@TheGregYang
@xai
@GoogleBard
@OpenAI
@krassenstein
@WholeMarsBlog
@Tesla
@WholeMarsBlog
restated the question. “Do you see
@xai
leveraging the custom silicon in the future”. “We are going to work with
@tesla
”, responds Elon, “it will have to be at arm’s length given the different shareholder base”. He acknowledges that it's a separate public company.
3
3
28
@karpathy
I did like Bing Creative more.
You didn’t copy someone else’s work. You used transfer learning.
1
1
27
@michael_nielsen
@gdb
Here’s what I’ve been using. A shortcut that uses whisper to transcribe the audio then GPT-3.5-turbo to make a summary, title and action items. You can run it directly from voice memos. Handles 30 minute clips in a minute or two.
2
2
25
@Iamtomblake
@nickfloats
@tldraw
A new tool by openAI lets developers build applications that take visual elements as inputs. In this case here,
@tldraw
has added a bit of logic to their general-purpose canvas app that takes the highlighted portion of your sketch along with any text or annotations and gives that
2
2
22
8
3
17
1
0
20
@vagabondjack
@_akhaliq
What about the objection that “trained” is misleading. The fine tuning is done on free and open data but the base model itself is actually using “pirated” data?
@sharongoldman
@mmitchell_ai
@databricks
The premise is a lie.
They only FINETUNED on an open source dataset.
The base model is GPT-J, which was trained on The Pile, which is essentially "dataset piratebay".
Like ChatGpt and ALL generative models, it is not fit for commercial use. It's also unsafe.
#CreateDontScrape
7
5
56
2
2
19
This is the paper mentioned in the
@thursdai_pod
talk this week
FreshLLMs: Refreshing Large Language Models with Search Engine Augmentation
paper page:
Most large language models (LLMs) are trained once and never updated; thus, they lack the ability to dynamically adapt to our ever-changing world. In this work, we
2
35
168
0
2
16
@yacineMTB
Okay. I’ll do it. Got any pointers that worked for you? Or should I pave my own way?
1
0
18
@eshear
It seems his disappointment is tied to the fact that he could have had what he wanted if he just bought the cake. That seems to be a big issue with dating after having so much freedom. It’s hard to transition to relying on others after doing it yourself. But that’s relationships
1
0
14
@detention361757
@WhoopDandy
@beala
@25thcenturygirl
@BobbyAllyn
It’s a quote from sam altman’s recent tweet
in regards to recent stuff about how openai handles equity:
we have never clawed back anyone's vested equity, nor will we do that if people do not sign a separation agreement (or don't agree to a non-disparagement agreement). vested equity is vested equity, full stop.
there was
1K
592
9K
1
0
14
GPT can use tools. This shortcut makes it easy to make and test them.
2
1
14
Although I will add that it has its own idea of what the rules are. But since it’s all pretend, I can ask it to clarify the rules and tell it when I don’t agree. I think it’s hilarious.
1
3
14
@eshear
I feel like asking what he’d prefer and then not doing it is worse than just not asking. What would be better still is if she had just let him know that she didn’t do what he asked.
2
0
14
@yacineMTB
I refer to it as a Система or a Програма which just mean system and program. These are both feminine. That’s in Ukrainian and Russian
1
0
12
Rabbit R1 unboxing
4
2
14
@alexgraveley
By many. I agree. But it seems ripe for long tail effects. Some guy is going to hammer that thing with more queries than the other 99% of people combined
1
0
14
WebXR assumes you have controllers for interaction. I made a modification so that you can respond to a pinch. Try it out here
1
0
13
Who else is generating cool avatars of themselves? I just made 103 for a friend and plan to reveal as a surprise. I’m posting a link because they offer a small reward and so others could try.
2
2
12
@ylecun
I’m loving the narrative by
@Wattenberger
around AI tools making us better writers ✍️
here's a quick write-up of a way I'm using
@OpenAI
's embeddings that I haven't seen before.
For my writing app, I want to highlight each sentence on any scale:
✨ happy — sad
✨ concrete — abstract
✨ + anything!
but had to get creative to do this cheaply, quickly, & robustly
33
47
401
0
1
12
@anammostarac
I think OpenAI is just paying to be the feature provider, similar to how Google has paid to be the default search engine. I’m glad there’s an approval step. I don’t like that typing into the search bar automatically sends a query to Google for search
1
0
11
The difference between this and what OpenAI demoed today is the “Prototype Shown” watermark at the beginning of the video
2
1
12
Running code with GPT? Freeze the libraries to versions that were available in 2021 for much better data rate!
2
2
13
0
0
13
@karenxcheng
@LumaLabsAI
I love how
@CorridorDigital
included this effect in their action NeRF with nerf
0
2
12
@icreatelife
I was going for something like, from school straight to the computer. The first one even looks like me from that time.
found webcam picture teenage boy playing on computer RuneScape mmo 2008 computer cramped tight space round glasses kitchen chair backpack --v 5
3
1
11
I enjoyed this. I enjoyed listening to the text of the MIT license.
AI-generated sad girl with piano performs the text of the MIT License
253
2K
8K
2
1
12