Visiting thoughts from an earlier time that felt similar to now in a business aspect. However similar, the learnings in this past year are knitted into my actions.
Observing the changes in the growth.
I wrote an article about pitching art businesses last year in September.
A pitch is a pitch, whether going after VC funds or grant funds or just for the sheer masochistic delight of it. Every pitch is an opportunity to gain clarity and focus.
Something has been missing from my story, and I have been looking in the wrong direction for that missing piece.
I suspected my lack of essential sales knowledge was holding me back from clarity in my pitch delivery. After all these years in tech, on the dork side of the road (happily), I have yet to fill my cup with sales and outbound marketing knowledge.
Sales is a money minting skillset in the right hands; the clever seller deftly crafts enticing narratives around anything that needs selling. A lack of swarthy sales maneuvering however, is not the missing component causing me to confuse rather than entrance the listener when I am in the pitch hot-seat.
Interlude: AI has become an exploration tool for my creative process. This is the latest in a series I am working on about geometric forms and expanding our thinking.
The latest pitch day was crafted by an incredible leader, who has been a beacon through the quagmire of tech layoffs of 2023 (Wendy, you are incredible). This remarkable leader, a dynamic woman with an impressive C.V. created a startup group for ex-techs who are ready to leave the corporate bonds for good, or at least entrepreneurially-curious.
The mega-Slack channel for laid-off tech expats has been a lifeline for myself and many in my world who were made "redundant." What an awful thing to say about a human in the midst of losing their livelihood.
I sound cynical here and perhaps I was. After the job I thought was going to be my dream job turned out to have the same and different corporate problems and blocks as other companies. I think it is important to acknowledge the people continually getting laid off from various jobs; we here about this all of the time.
There is no shame in getting laid off, and certainly, talk to a specialist if you find yourself struggling with mental health difficulties during and after a layoff.
If I can offer you one hopeful bright spot, a layoff can present a unique opportunity to deepen one's knowledge about oneself.
We've all had those moments where we wish we could unlearn something, right? Well, it turns out AI might be getting that ability, too.
Researchers in Tokyo have figured out a way to make AI models selectively "forget" specific information. Think of it as hitting the delete key for specific data points without messing up the whole system.
Why is this so cool? It could make AI way more efficient, protect our privacy, and even prevent those AI fails we all love to laugh at (but maybe shouldn't). They call this technique "black-box forgetting. "It's all about tweaking the instructions we give to AI, not the AI's inner workings. This is huge because it means we can apply it to those mysterious AI systems where the code is top secret.
Imagine an AI super focused on the task, doesn't guzzle energy, and can respect our privacy. That's the promise of this breakthrough. Read on for the highlights.
Problem Statement: Large-scale AI models like OpenAI's ChatGPT are versatile but inefficient for specific tasks due to their generalist design, which may also pose sustainability issues and privacy risks.
Innovation: The newly introduced method, "black-box forgetting," uses derivative-free optimization to modify the input prompts to AI models iteratively. This process allows the models to forget unneeded data without access to the model's internal architecture.
Methodology: The researchers employed the Covariance Matrix Adaptation Evolution Strategy (CMA-ES), an evolutionary algorithm, to refine prompts provided to CLIP, a vision-language model. They introduced a novel parameterization strategy, "latent context sharing," which simplifies the forgetting process and makes it scalable.
Results: This approach successfully made the CLIP model forget about 40% of the targeted classes through adjustments in how data was represented and processed without altering the model's core architecture.
Applications: The technique has significant potential to streamline AI models for specific tasks, making them more efficient and capable of running on less powerful hardware. It also offers a method to prevent AI from generating unwanted or harmful content and enhances privacy by enabling the removal of sensitive or outdated information.
Ethical Implications: This method addresses critical ethical concerns, such as the right to be forgotten, by providing a viable alternative to retraining models, which is often resource-intensive.
Reddit just launched a new AI tool called "Reddit Answers. "It's like having a super-smart research assistant sift through 20 years of conversations just for you.
Ask a question in plain English, and boom – the AI whips up a neat little summary of the most relevant discussions.
You can ask follow-up questions, dive deeper into specific threads, and get links to the juiciest conversations.
Purpose: Unlike other social media platforms that focus on content discovery, Reddit Answers is designed for efficient information retrieval from discussions dating back to 2005 across its 100,000 subreddits.
Functionality: The system processes natural language queries to generate curated summaries of discussions relevant to the questions posed by users. It offers a conversational interface that allows for follow-up questions, enhancing the user interaction with more dynamic search results.
Features:
Conversational Interface: Users can interact with the AI in a natural, conversational manner.
Curated Summaries: Provides condensed information from various threads.
Inline Snippets and Links: Direct links to related communities and detailed posts are included.
Follow-up Questions: Supports further inquiries to explore topics in depth.
Deployment and Accessibility: Initially available to a select group of U.S.-based English-speaking users, with plans t expand based on feedback and testing results.
Google's Project Mariner: Your AI Web Navigator
Google DeepMind just unveiled Project Mariner, an AI tool that's like having a super-efficient assistant for all your web-based tasks. But what exactly does it do, and how is it different from other AI agents we're seeing pop up?
Project Mariner in a Nutshell: It's designed to automate tasks within a web browser. Think about booking flights, researching products, or navigating complex websites. The key is that it combines language and visual understanding, so it can actually "see" and "interpret" what's on your screen, just like a human would.
Beyond Basic Agents: While other AI agents can follow simple instructions (like "find me a recipe for apple pie"), Project Mariner goes further. It can reason across multiple websites, understand complex tasks, and even respond to voice commands. This makes it more adaptable and capable of handling nuanced requests.
How it Works: It's built on Gemini 2.0, Google's latest multimodal AI model. This means it's been trained on a massive dataset of web pages and user interactions, allowing it to learn how humans navigate and use the web.
Key Differences from Other Agentic Tools:
Multimodal Understanding: It combines visual and language processing, making it more adaptable to different web environments.
Reasoning Across Websites: It can connect information from different sources to complete complex tasks.
Voice Interaction: It can understand and respond to voice commands, making it more user-friendly.
Project Mariner is still in its early stages, but it gives us a glimpse into the future of AI-powered web browsing. Imagine a world where you can accomplish online tasks with just a few spoken words or clicks. That's the kind of seamless experience that Project Mariner is aiming for.
AICharmLab - Your Creative Co-Pilot Leverage AI for more than just tasks—use it to spark your imagination.
What can I do with the AICharmLab app?
Design an empowering avatar as a reminder of how unstoppable you are.
Generate ideas for your personal or professional life with high-level specificity in a single prompt.
Generate perfect prompts you can reuse within AICharmLab or elsewhere in other AI tools.
Learn about AI through the interactive learning tool, print, and share your lessons with your team.
Generate task-specific templates and getting-started documents for anything you can think of, then visualize the idea!
Change your thinking or flip it upside down. With AICharmLab, you can work from images to text in your ideation process and flip back and forth, whatever works for you.
Was the AICharmLab app originally named TheFaceOfAI? —Yes, well, sort of. If you are new to the startup game, pivots happen at the large and small scale. Renaming things is the small pivot scale.
Complete this short survey to be added to the AICharmLab waitlist.
TheTechMargin LLC will never sell or share your data.
