Jeff Schneider
@jeffrschneider
Followers
1,599
Following
684
Media
387
Statuses
6,388
Explore trending content on Musk Viewer
FEMA
• 1643262 Tweets
North Carolina
• 1048501 Tweets
#turkishwomenareindanger
• 136699 Tweets
Best Record
• 121722 Tweets
Grammy
• 94825 Tweets
Pogba
• 80177 Tweets
ariana
• 69303 Tweets
Joker 2
• 68458 Tweets
#ikbaluzuner
• 49378 Tweets
Dolly
• 45100 Tweets
Flint
• 44263 Tweets
End of 1
• 23156 Tweets
Angers
• 17374 Tweets
Melo
• 16957 Tweets
Leeds
• 16115 Tweets
Magic Johnson
• 15815 Tweets
#SmackDown
• 13788 Tweets
UNLP
• 10609 Tweets
Four things caught my attention last week.
In no particular order:
1.
@AnthropicAI
released artifacts
Mostly hate it.
We implemented something similar in our chat app last year. The idea is to use a system prompt to catch specific requests to build tiny apps (like in html).
12
28
250
New arxiv rule: white papers that aren't peer reviewed must be written in Comic Sans font.
1
17
60
@yoheinakajima
Near real time audio streaming w/ASR interruption was the primary innovation.
So, improved versions of current stuff:
- IVR / telephony apps
- robot & kiosk interactions
- chat-to-device (PC, car, stuffed animal, ...)
1
3
44
@OpenAI
In 3 weeks you went from 'open' to 'partially closed' and 'non-profit' to 'for-profit'. Wow.
1
3
41
@DanielMiessler
@LangChainAI
Nah. It's pretty easy to pick up. A decent python engineer will be productive in under a week.
Great project, though!
0
0
34
@MarleneHenson12
@revrrlewis
Not sure it's just 'libs' ... it's normal people who think that she views an exponential mortality rate with levity.
0
0
28
Five things caught my attention last week.
In no particular order:
1.
@AdeptAILabs
splits the team.
A sign of the times.
Adept AI, who had raised over $415 million in capital, announced that their founding team were leaving and joining Amazon. The company would move their
3
5
29
My son and his bud covered some tunes in my basement. Enjoy.
2
1
16
After seeing the
@stanfordnlp
SQuAD adversarial results, it seems best to focus on distracting and misleading text. Have fun with the sample: Tweet your accuracy!
0
10
16
So, it's not "agents", it's:
APA: Agentic Process Automation
5
0
16
We're live!
In the next few months, we'll be focused on onboarding our first set of customers and ensuring their success. 🎯 The team will be hiring across all functions and raising fresh capital to expedite growth.
Give us a follow!
🚀 Excited to launch Imprompt, a new generative AI software company for the workforce! We automate tedious tasks with AI agents & offer a secure 'agent mesh' for enterprises. Meet the Imprompt Assistant, your business-focused AI 🤖💼
0
2
7
7
4
13
@yoheinakajima
@openai
announced more vaporware.
@metaai
crushed it with 3.1
@GroqInc
is demoing everything.
@SambaNovaAI
has a playground.
@langchain
declared everything is a tool.
Startups: born. death. repeat.
300ms voice is here.
LLMs are in the browser.
3
0
14
@robmay
Til then, it's probably best to not tweet headlines like "a year's worth of suicide attempts in 4 weeks".
Idea for your replacement tweet:
"1 Doctor, in 1 county, at 1 hospital indicates high suicide rate, but offers no actual data."
0
0
12
Thoughts on the AWS / Netflix outage, the
@ActiveState
response and my promise:
http://t.co/zojcbyXE
2
13
11
@BenjaminDEKR
Domain Specific Intelligence.
Go deep in one area.
Keep the model small, fast and cheap.
Allow it to be used by AGI as part of its composition of experts.
1
0
11
So many (young) people have no idea how hard Microsoft tried to kill Java.
3
2
11
It's final! Today is may last day at VMware. Hmm. I wonder what I'll work on next...
3
2
10
@Crimealytics
@hughw
Very interesting - but for so many of us who are red/green color blind, we had to ask someone else to interpret the colors. Quick fix is to change one of the colors to blue.
1
0
9
What if ...
Meta acquired Groq,
made their LPU technology open-architecture
similar to their Open Compute Project,
and funded the entire endeavor
by shorting Nvidia?
2
0
10
@mbusigin
I've had good luck with it, but I noticed it didn't do well on the Gorilla leaderboard:
0
1
9
Dropping a bomb on a Monday morning.
Function Calling is the wrong design center.
It's going to delay success in the AI agent and automation space by 2+ years.
Complete red herring.
10
0
9
People entering the NLP space should be required to sign a waiver.
"I am of right mind, and knowingly choose to inflict endless pain and suffering on myself, followed by brief but fleeting moments of hope."
1
1
9
Great quote from
@jasonbaldridge
"Don't get caught up in the cycle of getting papers out." Words to live by. 😉
1
2
9
Thanks to so many who directly or indirectly helped with ideas, code, conversation, and more.
@launchany
@LangChainAI
@transitive_bs
@hwchase17
@TheRealAdamG
@AtomSilverman
@appenz
@bentossell
@ericabrescia
@chrmanning
@syllogistic
@GregKamradt
@rachel_l_woods
@SullyOmarr
1
1
8
@PastorDScott
Only one network didn't focus on Puerto Rico (Fox News). Only one president focused on the NFL. The truth will set you free.
0
0
8
Why is everyone talking about Llama 3 and their benchmarks - and no one is talking about the 8K context window??
WTF am I going to do with 8k?
6
1
9
Each morning, I open my email to see if any of the waiting lists I'm on let me in. Sadly, each day is like:
- Google Bard API <-- NO
- Azure OpenAI <-- NO
- OpenAI GPT4 32k <-- NO
- AWS Bedrock <-- NO
- AWS Titan <-- NO
2
1
9
“The next generation of systems will be built using conversational interfaces,”... “This will be the interface to your systems. This will become the main interface to your systems. This is how you will unlock a much larger audience..."
0
7
9
This is an "unpublished" blog post, but it'll give you a feel for the level of effort put into trying to solve the problem.
Function calling is fundamentally flawed for "conversational dialog" use case.
@trung_rta
@BryanMcAnulty
3
2
8
I just published “Artificial Common Sense Reasoning”
0
4
9
I just found this in the code base.
ChatGPT is down AGAIN?
Status page shows every service including logins affected. Was down most of yesterday also.
What is going on?
17
1
26
1
0
9
A big milestone: we're moving to 'semantic composition' for adjective and adverbs that are gradable (aka, word scales). This will enable fine grained sentiment analysis.
1
2
9
One thing caught my attention last week.
It was Dare Obasanjo's post on Gen AI adoption, or the lack thereof.
He points to a Goldman Sach's report where several analysts comment on the lack of a killer use case.
Sadly, I agree with the report. Good use cases do exist
3
1
8
An amazing article on the war between craft beer makers and the mega-breweries:
0
0
8
1995 - JVM as a sandbox
1996 - EJB's as a sandbox
1997 - Servlets as a sandbox
2000 - VM's as a sandbox
2004 - SOA services as a sandbox
2006 - Cloud scaling as a sandbox
2013 - containers as a sandbox
2014 - FaaS as a sandbox
2015 - k8 as sandbox
2018 - all of the above, please
Nice thread comparing and contrasting JavaEE with today's arch patterns. Ah, the memories
0
0
2
1
2
8
I've been smelling this trash for 3 years, 291 days, 5 hours and 55 minutes. I think I'm going to sit outside for the next couple of hours with a cold beer, sipping - and waiting patiently for the garbageman to pick up the trash and dispose of it.
1
1
7
@Werner
I can not over-emphasize how important it is... that ever developer doesn't need to be a machine learning practitioner.
0
1
7
Oh, the irony. Watching humans write fake news about machines writing fake news.
0
0
7
K8/PKS is great, but the prize is what it enables.
VMware, after 5+ years of strategic moves, now has the building blocks in place to move aggressively on an As-a-Service strategy. Congratulations.
@PGelsinger
@shekar_ayyar
@cswolf
@pfazzone
@sakacc
0
0
7
Micro-release architectures. Micro-release architectures. Micro-release architectures. Repeat after me...
1
6
7
@SullyOmarr
that could be an interesting UI experience. you load the document in the chat interface; it scrolls & presents summaries, new prompts would be directed at the text on the screen, etc.
1
0
6
@CohereAI
Thoughts: Host it, create an API, turn it into an OpenAI compatible plugin, share it!
0
0
6
@emollick
ChatGPT taking acid, shrooms and smoking weed would do a better job than the top Gartner analyst.
Imagine being a Gartner analyst, and never, and I mean NEVER, having looked at the software you were covering.
2
0
6
Mom: why did you kick your sister?
Kid: Because my attention was on her.
Mom: be more specific.
Kid: Because of the vector.
Mom: huh?
Kid: See embedding, 122, 01, 13, 23, 199, 42, 11, 28, 221, 49, 11, 68, 188, 202, 11, 109, 74
Mom: Don't you MAKE ME decode that vector, kiddo!!
0
0
7
@DivGarg9
You thinking about running it in the cloud in 'desktop as a service' mode? Slap an API on it...
2
0
7
@estebrunette
@TwitchSimba
@JoshHelaku
@neiltyson
Actually, the word you attempted to spell is "hundredths", and that is the incorrect usage. It is used for in decimal numbers to identify portions of a hundred. "It lasted for five one hundredths of a second."
2
0
7
American's want a president that isn't bat-shit crazy, a congress that works together, and a press that reports it fairly.
0
1
5
@hwchase17
Here's the best part... you don't even have to run the code.
1. Take a screenshot of your code (minus the answer) and add it to a GPT4 vision prompt
2. Use a prompt like: "imagine this program and how it works; simulate running it and show me the output."
Twitter
⛏️🦜 New extraction guides 🦜⛏️
Reliably extracting structured data from unstructured text is one of the killer use-cases for LLMs. It's a fantastic way to bridge the gap between LLMs and traditional APIs and systems.
We've put a lot of work on this lately, including an open
3
56
315
0
2
7
@haldaume3
A smart machine would determine the delta between archetype, prototype & stereotype.
0
1
7
As 'elastic' becomes the norm, the focus will shift to 'resilience'. From 'elastic cloud' to 'resilient platforms'...
1
15
7
@SullyOmarr
Things you got right:
✓ Tabular Interface
✓ Generative UX
✓ Constrained Agents
✓ Practical Templates
✓ Simple PLG Pricing
Things you got wrong:
- nothing.
1
1
10
@sandhillstrat
A 4,000 person I.T. shop faces different problems than a startup. Different approaches. Different solutions.
1
0
6
I'll take a win on a Friday afternoon.
I'm calling these "Natural Language API's" (NL-API).
OpenAI Plugin compatible.
For now, we only have No-Auth, but hope to have all authorization done next week.
Thanks
@vercel
@transitive_bs
@OpenAI
1
0
6
@fchollet
Seriously, no one called bullshit on this? "The intelligence of a human is specialized in the problem of being human." - classic.
1
0
6
The first in our new series.
#SupportRealActors
Think AI's here to snatch your job? 🚫🤖
AI's not your replacement but your new sidekick to vibe with the tasks you're not into. 📊 Upskill & take on what you love with Imprompt.
#AI
#WorkSmart
#ImpromptInnovation
0
2
4
2
0
6
Can someone please...
build an API that assesses the usability of a website on mobile.
- Use agents with nav and image analysis.
- Return overall assessment & recommendations
3
0
6
Just returned from the Google Next Conference. I made a point to have 100 conversations w/attendees.
- Gen AI was everywhere
- Many people think Gen AI is hype, and still compare it to blockchain.
- The new architecture hasn't emerged; some talk about Gen UI and GenOps.
3
2
6
@garrytan
also, some 'fails' are soft, like you see in RAG, while API calling / tool use are almost always hard fails.
1
0
6
<-- go here
Pick a character, but try not to get sucked into the personality.
Just talk with it (mic on) and test it.
Put it in its own window. Do a long session, like 5 to 10 minutes. Keep it running in the background while you surf the web, etc.
3
1
6
Pumped up. Thinking about quitting artificial intelligence and moving to El Paso to build a wall, pick vegetables and arrest immigrants!
0
3
5
@hwchase17
@yoheinakajima
@tb_tomaz
@neo4j
24 years later, Emil continues to reinvent neo4j.
Harrison - just remember this conviction for LangChain in the year 2048. 😀
1
0
6
start with a domain name
and work your way to a product.
that's what they teach in business school, right?
1
0
7
@SullyOmarr
so that crazy system prompt is used in their API.
hmm.
seems like you could copy that system prompt over to Groq, etc. and get similar capabilities.
1
0
6
#AWS
sees new pattern: if central service loses state, reverse engineer state from provisioned entities:
http://t.co/hLSvw92w
< yup.
1
4
6
Interesting thread. It indicates that the 'uncanny valley' theory is either incorrect, or doesn't apply to 'artificial conversations'. The negative reactions focus on 'intentional deception' by the bot makers (which is the entire point of an artificial role-based doppelgänger)
Google Assistant making calls pretending to be human not only without disclosing that it's a bot, but adding "ummm" and "aaah" to deceive the human on the other end with the room cheering it... horrifying. Silicon Valley is ethically lost, rudderless and has not learned a thing.
278
3K
8K
0
1
6
@hwchase17
This is perhaps one of the most important decisions that'll be made in this space: how do we partition memory, as to not overshare, violate privacy, ensure policies on retention, etc.
1
0
6
@robmay
That's a tiny sample population. Just a quick scan of all English news articles over the last year indicates that references to suicides were highest in September of 2019, while last 3 months have averaged lower than the 2019 mean. It'll be interesting to see real data.
1
0
6
"The beauty and tragedy of the modern world is that it eliminates many situations that require people to demonstrate a commitment to the collective good. Protected by police and fire departments and relieved of most of the challenges of survival an urban man might go through...
1
0
5
Mandatory reading for Voice UX designers: "Some Universals in Language Usage" by Penelope Brown and Stephen Levinson, "The Pragmatics of Politeness" by Geoffrey Leech, and "Questions and Politeness" by Esther Goody.
The case against teaching kids to be polite to Alexa
1
1
2
0
2
5
@SullyOmarr
gpt4 vision did a decent job of getting the numbers:
and I cheated a bit to get the plot (w/out using code interpreter):
2
0
4
This is the moment where we can finally focus on mining knowledge; it took two stinking years to build the framework... but we're here. Sigh.
1
0
5
@robbensinger
@OpenAI
OpenAI should just publish their actual governance policies. Proper 501(c)(3) governance ensures that R&D isn't quietly fed to private ventures; shows process for full disclosure. IMO, the non-binding charter is a bit silly.
0
0
3