Interesting approach to spotting #LLMs in papers. Not sure it’s a problem though: I’d actually be shocked if rationale minds didn’t use an AI assistant to speed things up. Of course the question is who’s doing the final editing…
scientificamerican.com/artic…
Guillaume Decugis retweeted
How to be as "smart" as Auto-Regressive LLMs:
- memorize lots of problem statements together with recipes on how to solve them.
- to solve a new problem, retrieve the recipe whose problem statement superficially matches the new problem.
- apply the recipe blindly and declare victory.
- do not use basic logic.
- do not use common sense to check your solution.
- do not use a mental model of the situation as a sanity check.
- do not simulate the scenario in your mind using your world model.
- when someone tells you your solution is wrong, reply "I'm sorry, you are right" and apply another irrelevant recipe.
Knowledge accumulation is not a substitute for actual understanding.
Guillaume Decugis retweeted
This is the first demo of any agent, leave alone coding, that seems to cross the threshold of what is human level and works reliably. It also tells us what is possible by combining LLMs and tree search algorithms: you want systems that can try plans, look at results, replan, and iterate till success. Congrats to Cognition Labs!
Today we're excited to introduce Devin, the first AI software engineer.
Devin is the new state-of-the-art on the SWE-Bench coding benchmark, has successfully passed practical engineering interviews from leading AI companies, and has even completed real jobs on Upwork.
Devin is an autonomous agent that solves engineering tasks through the use of its own shell, code editor, and web browser.
When evaluated on the SWE-Bench benchmark, which asks an AI to resolve GitHub issues found in real-world open-source projects, Devin correctly resolves 13.86% of the issues unassisted, far exceeding the previous state-of-the-art model performance of 1.96% unassisted and 4.80% assisted.
Check out what Devin can do in the thread below.
Too many LLMs?
An Annotated and Updated Directory of GenAI Models: docs.google.com/spreadsheets…
Guillaume Decugis retweeted
The best founders I know — no matter their company’s scale — thrive on doing customer support directly. There’s literally no better way to understand the pulse of your customer base, what features to build next, or where systems are breaking down. It’s always upside.
Good summary by @VentureBeat @mmarshall of why open-source LLMs are getting traction. With concrete examples too: venturebeat.com/ai/how-enter…
Guillaume Decugis retweeted
Remember when GPT-2 was deemed too dangerous to release?
That was 5 years ago.
The world didn't end.
In fact nothing bad happened.
[Yes, I made fun of them then]
Guillaume Decugis retweeted
AI providers need a clear commitment to safety, unquestionably. But for enterprises —the real paying customers— to trust the AI that they’re incorporating into their businesses, they’re going to need ironclad guarantees that the AI models one day won’t disappear or be altered because a few people think it should. We’re going to need a new era of charters for AI model providers if AI is going to reach any form of mainstream.
Guillaume Decugis retweeted
Personally I am short chat interface and long on SaaS and Enterprise implementations that use LLMs
These new startups will perform extraordinary tasks that make a big difference in lowering cost or solving previously unsolvable problems and create new markets in the process
Guillaume Decugis retweeted
Riding the wave of video content surge on social media, we've taken a significant step in expanding our product offering. We unveiled our video analysis solution at #meltwatersummit to navigate the complex, multidimensional realm of videos, proficiently executing tasks like Logo Detection, Celebrity Recognition, Object Detection, and Scene Detection. By effectively tackling temporal dependencies and variations in lighting, scale, viewpoint, and occlusion, we're enabling businesses to extract profound, actionable insights from dynamic video content.
When coupled with our existing Image Analysis and advanced Natural Language Understanding capabilities, this development further fortifies our platform as a robust tool for data-driven decision-making and marketing optimization.
Image (and video) recognition is getting increasingly good.
Will that create a market for adversarial fashion (clothes that trick the tools into thinking you're something or someone you're not)? Hopefully, because I'm easily amused. #MeltwaterSummit
Guillaume Decugis retweeted
“Marketers won't be replaced by Al. Marketers not using Al will be replaced by marketers using Al.”
@Meltwater #MeltwaterSummit
Guillaume Decugis retweeted
Meltwater’s Head of Consumer Insights, Guillaume Decugis, in conversation with @MoetUSA @LVMH’s @L_Spencer on building a transformational Social Intelligence Program. What a brand, what a conversation! #MeltwaterSummit
Discovered yesterday at the #MeltwaterSummit that @Trevornoah knows a lot on #AI (Document-level vs entity-level sentiment, anyone? ;-)) Great to have his pragmatism, critical thinking and common sense on this topic that has become very passionate these days.
Guillaume Decugis retweeted
Soundbite of the day by @Trevornoah: "There is no guarantee of success but there is definitely a guarantee of unhappiness if you choose to work with assholes." #MeltwaterSummit @Meltwater #hardtruths
Full house as we kick of the #MeltwaterSummit Amazing speakers and a ton of innovation launches. Looking forward to an exciting event over the next 2 days.
Guillaume Decugis retweeted
8 years. and people think AGI imminent? 🙄
This Post is from an account that no longer exists. Learn more
Simply slapping on AI (or even Generative AI...) won't guarantee success for a product.
Here's a good read if you already start to feel you've been pitched AI too many times these days (designed for investors but I think it works if you're a potential bu…lnkd.in/eY82gbbi