#OpenAI: "We’re planning to release our first open language model since GPT‑2 in the coming months. We’re excited to collaborate with developers, researchers, and the broader community to gather inputs and make this model as useful as possible. If you’re interested in joining a feedback session with the OpenAI team, please let us know below." #LLM #AI #MLopenai.com/open-model-feedback…
😲 DeepSeek-V3-4bit runs at >20 tokens per second and <200W using MLX on an M3 Ultra with 512GB. This might be the best and most user-friendly way to run DeepSeek-V3 on consumer hardware, possibly the most affordable too. You can finally run a GPT-4o level model locally, with possibly even better quality. #LLM #AI #ML #DeepSeek #OpenAI #GPT #OpenWeight #OpenSourceventurebeat.com/ai/deepseek-v3…
DeepSeek's free 685B-parameter AI model runs at 20 tokens/second on Apple's Mac Studio, outperforming Claude Sonnet while using just 200 watts, challenging OpenAI's cloud-dependent business model.
Thing I want to exist but don't want to build: a #mud that does pretty much the same thing as #SillyTavern or #agnaistic, intended for selfhosting. A MUD would be an excellent interface for #AI assisted #roleplay. With #Evennia and #OpenRouter, this honestly wouldn't be that hard. I just hate #Python and don't wanna.
The way to do it would be: 1. Set up a #docker#Evennia environment, to make distribution easy. 2. Add an attribute on Evennia accounts to hold an openrouter API key, and a userflow to get a users key via OAuth: openrouter.ai/docs/use-cases/oauth-pkce3. Use the #OpenAI Python API and hook it up to NPC's to let them chat, generating responses with the user's API key. Figure out function calling and hook it into the various creation functions of Evennia for extra fun. Probably need to store attributes for what models and parameters to use on each NPC, with global defaults probably from the user's account. 4. Dump the room name and desc, the player desc, and chat history into the AI context. 5. Something something MSP/MXP and ElevenLabs for sound effects and music. 6. Something Something MXP for images. 7. Dockerize the thing and upload it to let users run there own muds for AI roleplay for themselves and their friends.
If I either liked Python or was unemployed, I would do this. If only AI coding was good enough to just make the thing based on my design skeleton above.
Techmilliardär Musk will zusammen mit Investoren OpenAI kaufen - für fast 100 Milliarden Dollar. OpenAI-Chef Altman, ein Musk-Erzfeind, reagiert mit Spott. Auch der Verwaltungsrat sage Nein. Von. R. Spiegelhauer.
Zkousim novy research mod od OpenAI. V zasade je na tom videt, jak jsou vsichni hraci (slovo velmi hodici se) v reaktivnim modu. V jednotkach dni po sobe takticky zverejnuji male pokroky.
Research: 🤷 Efektivne likviduje agent startupy nad jejich API 💸 Jeden dotaz sezere ~500k tokenu => $10 dolaru z ceniku API 🙅 Ignoruje nastaveni robots na webu, kde hleda (kdo by to byl cekal)
Ale musim rict, ze vysledky sou super. Za par drobnych propalite VC fondum $1k elektriny mesicne.
Son muchos quienes hablan del "sesgo ideológico" de la AI china #DeepSeek No les muestren este prompt y respuesta, que explican como funciona Bing Image Creator que utiliza #DALL-E, desarrollada por #OpenAi desde la "cuna de la libertad occidental".
Ahora que los AI-bros de Silicon Valley están panicando con la salida de #Deepseek y sus capacidades a la par de las de #OpenAI aunque a una fracción del precio, además de ser en buena parte open source, una crítica habitual al modelo es que es chino y, entre otras cosas, te censura cualquier mención a los sucesos de la Plaza de Tiananmén en 1989.
Pues vamos a ponernos empíricos y hacerle la pregunta en local para ver qué hay de verdad. Nos bajamos la versión 14b del modelo y le pedimos un resumen de esos hechos.
Oh, sorpresa. Habla de represión, de censura y de una auténtica masacre.
El modelo no será perfecto y tendrá sus sesgos como todos. Pero criticarlo por ser chino no se sostiene.
People are seriously criticizing the integration of whisper into VLC? It's an accessibility feature that will benefit everyone, and with little impact if not perfect. It runs on device and is exactly the type of improvement we'd expect from reasonable AI systems. The only issue I see is that the training data isn't open source, but the model and inference code is. #VLC #whisper #OpenAI
What is this? A company valued at $157B doesn't know how to produce a demo with decent audio. I don't even expect great quality, but this is just crap. #OpenAIyoutube.com/live/XKABimtOWME
I was just talking to a colleague about the AI bubble. These companies are in so deep they can't tell the truth. They are all lying about the efficacy, costs to consumers and most importantly how & when this tech works or doesn't.
Is there enough money on the line to kill over?
There's likely a trillion bucks of valuations across the industry. Billions in sunk costs, billions in c suite remuneration, billions in VC mgmt costs.
On the sixth day of OpenAI's Shipmas, a notable enhancement graced ChatGPT—the integration of camera access within its Advanced Voice Mode. This feature,
So a crazy thing happened. In a crime thriller film called Les chambres rouges (Red Rooms, 2023) there is a scene where the hacker protagonist is attempting to purchase a snuff film in online auction. The auction happens via IRC or IRC-like chatting environment and to my surprise there's actually me @rolle and my wife @mustikkasoppa who are one of the bidders.
1) The nickname of my wife is mistakenly written as "mustikasoppa" (with one k) but if I recall correctly she has used a mistakenly written nickname in the past 2) We are both operators and on the same IRC channel as we've been for the past 17 years
This is not a coincidence. One nickname can be made up by accident but not two with these features and definitely not in a French movie. My wife's nick "mustikkasoppa" is Finnish and means "blueberry soup".
Our IRC logs are more or less public because of open source and statistics so I presume the scene has been made with chatgpt which has scraped our nicks from the Internet so that they ended up in the movie. We still chat via IRC every day together.
The era of AI... Do your background check, folks. I'm glad this wasn't a dramatized documentary film but a complete fiction. However my wife and I are real. Mixed feelings.
😲 OMG! Mlx-whisper transcribed 12 minutes of speech under 18 seconds with excellent accuracy, using the new OpenAI model, whisper-large-v3-turbo, on my MacBook Pro with the M3 Max! ⚡ #OpenAI #ML #AI #Transcriptionhuggingface.co/mlx-community/w…
Independent test of #OpenAI’s o1-preview model achieved near-perfect performance on a national #math exam (landing in the top .1% of the nation’s students).
o1 also outperformed 4o on the math test, but took about 3 times longer to do so (10 minutes vs. 3 minutes).
OpenAI launches new AI model family o1 claiming PhD-level performance venturebeat.com/ai/forget-gpt-… (I don’t seem to have it yet in my Plus account) #AI #OpenAI
Good. Biden's appointment of Altman to the AI safety and security board is up there with his Garland and Fauci appointments. Bad decisions.
The man has been accused by those who know him in his own inner circle of being "not consistently candid" and if there's anyone who should not be deciding AI risk issues it's someone not candid.
"Scarlett Johansson scandal puts OpenAI's Sam Altman back in the hot seat"
“I have seen the extremely restrictive off-boarding agreement that contains nondisclosure and non-disparagement provisions former OpenAI employees are subject to. It forbids them, for the rest of their lives, from criticizing their former employer.”
Like many other technologists, I gave my time and expertise for free to #StackOverflow because the content was licensed CC-BY-SA - meaning that it was a public good. It brought me joy to help people figure out why their #ASR code wasn't working, or assist with a #CUDA bug.
Now that a deal has been struck with #OpenAI to scrape all the questions and answers in Stack Overflow, to train #GenerativeAI models, like #LLMs, without attribution to authors (as required under the CC-BY-SA license under which Stack Overflow content is licensed), to be sold back to us (the SA clause requires derivative works to be shared under the same license), I have issued a Data Deletion request to Stack Overflow to disassociate my username from my Stack Overflow username, and am closing my account, just like I did with Reddit, Inc.
The data I helped create is going to be bundled in an #LLM and sold back to me.
In a single move, Stack Overflow has alienated its community - which is also its main source of competitive advantage, in exchange for token lucre.
Stack Exchange, Stack Overflow's former instantiation, used to fulfill a psychological contract - help others out when you can, for the expectation that others may in turn assist you in the future. Now it's not an exchange, it's #enshittification.
Programmers now join artists and copywriters, whose works have been snaffled up to create #GenAI solutions.
The silver lining I see is that once OpenAI creates LLMs that generate code - like Microsoft has done with Copilot on GitHub - where will they go to get help with the bugs that the generative AI models introduce, particularly, given the recent GitClear report, of the "downward pressure on code quality" caused by these tools?
While this is just one more example of #enshittification, it's also a salient lesson for #DevRel folks - if your community is your source of advantage, don't upset them.
Europäische Datenschützer reichen Beschwerde gegen OpenAI ein
Das KI-Programm ChatGPT veröffentlicht Daten über Privatpersonen, deren Herkunft unklar sei. Das werfen Datenschützer dem Anbieter OpenAI vor. Zudem ließen sich falsche Angaben nicht korrigieren.
Someone on LinkedIn pointed out that #OpenAI doesn’t use an LLM for their support chat, but has instead opted to use some extremely limited decision tree. Which is both understandable, telling, and fucking hilarious.
Interesting #ChatGPT prompting technique! 1. Telling #AI model to “take a deep breath” causes math scores to soar in study! 2. #OpenAI DALL-E pays more attention to words in all caps. 3. Don't forget to say thank you and please to #LLM because Simon Willison thinks "in the training data, there are lots of examples where a polite conversation was more constructive and useful than an impolite conversation." arstechnica.com/information-te…
OpenAI threatening to leave the EU if they pass legislation requiring them to list their data sources because: "In addition to the possible business threat, forcing OpenAI to identify its use of copyrighted data would expose the company to potential lawsuits. Generative AI systems like ChatGPT and DALL-E are trained using large amounts of data scraped from the web, much of it copyright protected. When companies disclose these data sources it leaves them open to legal challenges."
Like every other boom tech company, it's just doing labor crimes and theft and claiming you're actually too innovative to be regulated, God.
OpenAI CEO Sam Altman has warned that the company could pull its services from the EU if it finds upcoming regulations too onerous. The EU AI Act is currently being finalized by lawmakers and should become law next year.
ChatGPT has leaked conversations and payment info of some users. While #OpenAI feels "awful" about this, experts say users "should have had zero expectation of privacy when using the #ChatGPT web demo."
Reading about #ChatGPT plug-ins and wondering why this is framed as plug-ins for #ChatGPT (giving it "capabilities") rather than #ChatGPT as a plug-in to provide a conversational front-end to other services.
Nevermind, I know why: This is #OpenAI yet again trying to sell their text synthesis machine as "an AI".
In most movies that feature time travel, traveling back in time is instantaneous or takes just a few seconds, regardless of how long ago a person travels back. The exceptions I know of are two of my
"Here’s the thing, though: the system will frequently give the wrong answers * the real skill in the homework assignment will be in verifying the answers the system churns out — learning how to be a verifier and an editor, instead of a regurgitator.
What is compelling about this new skillset is that it isn’t simply a capability that will be increasingly important in an AI-dominated world: it’s a skillset that is incredibly valuable today. "
The first obvious casualty of large language models is homework: the real training for everyone, though, and the best way to leverage AI, will be in verifying and editing information.
#Whisper is an #openSource #speechRecognition model written in #Python by #OpenAI. I’ve just seen it in action. Extract an #mp3 from a video, run it through Whisper, and it turns every spoken word into text. It even does a very decent job in #Danish. Perfect for subtitling #TV and #video. I am very impressed.
