
Sign up to save your podcasts
Or


Show Notes
* Chatbot PI's New Update - The chatbot PI has released a new update for their iOS app where users can now speak to it. The bot will talk back, allowing for more interactive conversations. The update also includes six different voice options. However, the feature has not been tested yet due to hardware limitations. [Timestamp: Start - 5:00]
* GPT API Release - The long-awaited GPT API is now available for everyone to start experimenting with. The API doesn't seem to have any drastic price cuts, but it's expected that more people will start integrating it into their workflows and apps. [Timestamp: 5:01 - 10:00]
* OpenAI's Super Alignment Team - OpenAI has announced the creation of a super alignment team. The team will be dedicated to steering AI systems that are much smarter than humans. They aim to build AI aligners that can do a better job than human alignment teams. OpenAI is dedicating 20% of their compute to this effort. [Timestamp: 10:01 - 15:00]
* Public.com's New Co-pilot - Public.com, an investment app, is building a co-pilot into their trading app. The co-pilot will be able to provide customized insights relevant to your portfolio and your goals. The feature will be rolled out in three phases, with the first phase already available. [Timestamp: 15:01 - End]
* Try out PI's new iOS update and share your experience in the comments.
* Check out the GPT API and explore its potential applications.
* Follow OpenAI's super alignment team's progress and understand their efforts.
* Test Public.com's new co-pilot feature and see how it can help with your investments.
Closing Remarks In this episode, we looked at PI's new iOS update, the rollout of the GPT API, OpenAI's super alignment team, and Public.com's new co-pilot feature. Stay tuned for more updates in the world of AI. See you next time!
Links:
https://twitter.com/heypi_ai/status/1676998045567680514
https://openai.com/blog/introducing-superalignment
https://public.com/alpha
Hosted on Acast. See acast.com/privacy for more information.
* Stable Diffusion 0.9: A significant release from Stability AI. The new model has impressive demo images, including an alien walking through Las Vegas and a wolf in Yosemite National Park. Notably, the model has made significant improvements in rendering hands and faces.
* ClipDrop.co: A toolset from Stability AI where you can experiment with the 0.9 model. Examples include a lion with wings, a lighter cake, and a farmer cat in a garden.
* Image Generation: Discussion of various generated images, including a scene inside a glass bottle and a hand holding a heart.
* Adobe Express: Similar to Canva, Adobe Express is a straightforward, templated design tool. It now incorporates Adobe Firefly, allowing users to generate images directly into their workflow.
* Color GPT: A simple tool for generating color palettes based on user prompts.
* Scribble Diffusion: A fun program that turns sketches into images.
https://clipdrop.co/stable-diffusion
https://stability.ai/blog/sdxl-09-stable-diffusion
https://scribblediffusion.com/
https://new.express.adobe.com/
https://www.colourgpt.app/
Hosted on Acast. See acast.com/privacy for more information.
OpenAI API Updates
* Notable reduction in the cost of some models:
* 75% cost reduction on embeddings
* 25% cost reduction for input tokens for GPT 3.5 Turbo
* Function Calling Upgrade for GPT3 and GPT3.5:
* This new feature allows conversion of natural language queries into API calls or database queries.
* Examples include converting queries like "Email someone to see if she wants to get coffee on Friday" to function calls such as "send an email."
* It can even convert database queries like "What are my top 10 customers this month?" into API calls like "get customers by revenue."
* More complex examples involve using this feature to answer questions or pull specific data.
* Excitement for the new bigger context windows which would allow for new applications.
OpenAI API Updates
Facebook and Meta's New Generative AI Model
* Facebook and Meta have developed a generative AI model for speech with state-of-the-art performance.
* This model seems to be a major advancement in voice generation.
* Check out the linked 2-minute video for an overview of this new technology.
Facebook and Meta's New Generative AI Model
AI Speech Classifier from Eleven Labs
* A tool that allows users to upload a 60-second audio file to detect if the speech was AI-generated.
* Discussion on the potential benefits of these types of classifiers, particularly in the early stages of AI development and use.
* Speculation about future tools for detection of AI-generated content, including potential metadata requirements for certain photos.
AI Speech Classifier from Eleven Labs
Hosted on Acast. See acast.com/privacy for more information.
Welcome back, everyone, to our latest deep dive into some of the most interesting technology and design tools out there. Without further ado, let's delve straight in.
Music Gen: A Novel Music Creation Tool
Our first highlight is an impressive demo from Music Gen, a creation of Facebook's research group. Similar to Google's LLM, Music Gen facilitates music creation with a unique twist - the ability to condition melodies. Despite a few connection hitches during our demo, it eventually lived up to expectations, crafting a 90s rock song with heavy drums and an electric guitar, just as we had specified.
Not only does it give you several pre-filled prompts to work with, but you also get the chance to experiment with creating your own tunes. The more I explore it, the more I realize how much potential there is for non-musicians like me to tap into their latent creativity. With Music Gen, music production is becoming increasingly accessible, and we look forward to seeing how this evolves.
ClipDrop by Stability AI: Powerful Image Manipulation
Next up is ClipDrop by Stability AI, a site teeming with valuable image manipulation tools. Whether you're looking to replace a background, upscale an image, remove unwanted elements, or clean up your images, ClipDrop has got you covered.
A standout feature is the uncropping tool, quite similar to Adobe's Generative Fill, which seamlessly fills in the edges of an image. It performed remarkably well even on abstract art, offering a broader view of the initial image. Furthermore, their Re-Imagine XL tool provides various alternatives from a single image - perfect for when you need a fresh perspective.
While ClipDrop's free account has some limitations, like watermarked images, the range of offerings in their paid version is worth considering, especially if it becomes part of your workflow.
Bard's Continuous Improvements: Google Sheets and More
Bard recently rolled out some understated updates that further enhance its user experience. They now provide the option to export data directly to Google Sheets, a feature that data analysts might find particularly useful.
Bard also upgraded its tech computational prompts, improving its ability to tackle mathematical tasks, coding questions, and string manipulations. They run code in the background using Python, akin to ChatGPT's code interpreter plugin. While there is no option for file uploads yet, this step towards executing code on behalf of users indicates a promising trajectory.
Automator for Figma: Design Automation at its Best
Last but certainly not least is Automator for Figma. This plugin allows you to automate tasks such as spec'ing out your design or importing data from Airtable, saving you valuable time. It even comes with a find and replace feature.
What's more, the community page offers user-created automations, providing inspiration on how to best utilize the plugin and streamline your design process.
That's it for this week's exploration of tech and design tools. We hope you found these insights helpful and feel inspired to check out some of these impressive tools. Stay tuned for more exciting updates in our next piece. Thank you for joining us, and we'll see you next time!
https://huggingface.co/spaces/facebook/MusicGen
https://clipdrop.co/
https://bard.google.com/updates
https://automator.design/
Hosted on Acast. See acast.com/privacy for more information.
Welcome back to another exciting installment with me, John Siwicki. This week, we have an eclectic mix of fascinating things that caught our attention. Let's dive straight in!
First up, we have some intriguing updates from Bard. A couple of days ago, Bard started incorporating location information into their searches, promising a leap towards more localized search results. However, it's important to note that Bard still labels this feature as an experiment.
In a test search, we asked Bard to find a place for an oil change - a standard request, right? The first results interestingly pointed us to Colorado, despite our location being elsewhere. Clearly, Bard's location feature has some kinks to iron out, but the potential here is undeniable. The local search results, complemented by added images about oil changes, show promising strides. We eagerly await future improvements to Bard's location-based services.
Next, we'll shift gears and explore the ChatGPT plugin. Early on, many of these plugins felt somewhat nebulous, with unclear functions. However, a recent discovery caught our eye: the 'Show Me' plugin. This fantastic tool shows you how certain files work. For instance, it vividly demonstrated how a CSS file operates within a web browser. An added bonus? You can edit the diagrams it generates, providing hands-on learning!
Another exciting development is the recent addition of a search feature to the ChatGPT plugin store. Now, finding plugins like 'Show Me' is easier than ever. This proves immensely helpful for your charting needs.
Moving on, we discuss the impressive work by po.com, a site run by Cora. It provides a unique platform for testing various language models, even those typically challenging to access. One such feature is the 'Mid Journey Bot,' which assists you in writing more engaging mid-journey prompts.
Here's an example: we entered a prompt about a house in Miami after a storm, with a child in a raincoat in the foreground. The bot expanded upon this simple scene, weaving a detailed and evocative narrative. The depth it added to the original prompt was truly remarkable, hinting at the immense creative potential of such tools.
Lastly, we're excited to introduce you to an intriguing app discovered on Twitter: Audio Pen.ai. This innovative platform offers an easy way to convert your spoken thoughts into structured text. Record your ideas into the app, and it organizes them into a clear and coherent flow. We found this tool especially helpful for managing voice notes.
That's it for this week! We'll continue to explore these promising developments and share our findings with you. We hope you found these insights as exciting as we did, particularly the potential of Audio Pen. Remember, we'll include all relevant links in the show notes.
See you all next week. Thank you for joining us!
Links:
https://poe.com/Midjourney
https://audiopen.ai
https://bard.google.com/updates
Hosted on Acast. See acast.com/privacy for more information.
We kick things off by dissecting Microsoft's latest innovation, Windows Co-Pilot. We delve into its range of functionalities, such as interacting with system settings, summarizing PDFs, and initiating Spotify playlists. Listen in as we shed light on its potential capabilities and anticipated impact on user experience.
Then, we turn our attention to Adobe's newest beta update for Photoshop - the Generative Fill. Accompany us through a detailed walkthrough of this revolutionary feature, which utilizes text-based commands to modify images. We contemplate its potential to transform the landscape of graphic editing.
Wrapping up the episode, we discuss a novel feature in ChatGPT, enabling users to share chat links. We explore how this unique innovation may alter our engagement with this AI chatbot, potentially introducing a fresh dimension of sharing and interaction.
https://www.adobe.com/products/photoshop/generative-fill.html
https://blogs.windows.com/windowsdeveloper/2023/05/23/bringing-the-power-of-ai-to-windows-11-unlocking-a-new-era-of-productivity-for-customers-and-developers-with-windows-copilot-and-dev-home/
Hosted on Acast. See acast.com/privacy for more information.
Links:
https://openai.com/blog/introducing-the-chatgpt-app-for-ios
https://aitestkitchen.withgoogle.com/experiments/music-lm
https://zoo.replicate.dev/?id=a-comic-book-panel-of-ocean-waves-pixel-art-by-salvador-dali-lpe55wlg
Transcript:
Hello, everyone. Welcome back to the show. I'm John Siwicki, and I appreciate you joining us on our weekly AI roundup. Today, we're going to explore three intriguing developments. A few of these are quite entertaining as well.
Let's begin with something recent: the release of the Chat GPT iOS app. It's the first version, and an Android version is not available yet. The app has a few interesting features. It integrates with the Whisper API, which ensures excellent text-to-speech capabilities. Additionally, if you're a Plus subscriber, you can use GPT-4 on it, which is an impressive feature.
Currently, we're hoping for an Android app soon. I'm curious to see if a well-crafted mobile app will influence user habits and possibly spark new use cases. I've used the web app on my mobile device; it got the job done, albeit not smoothly.
Therefore, it will be fascinating to observe how a well-designed mobile app changes user behavior. Will people start using it more frequently? Will they opt for it over Siri or Google Assistant? The next few weeks will certainly provide some insights, especially if the app evolves and introduces new features.
Moving on to the second topic: Google's new AI test kitchen. One of their projects is Music LM, which enables text to audio conversion. I tried one demo, inputting "80s synth music that I can listen to on a run after a breakup." I think it's a fascinating tool, especially for those who aren't musically talented. I've been experimenting with it all weekend, and I recommend getting on the waitlist if you can.
Lastly, we'll discuss a text-image playground by Replicate called Zoo dot replicate dev. It lets you experiment with different text-image combinations. For listeners, I used a comic panel of "Ocean Waves Pickle Solar" by Salvador Dali and compared the results by different models. It's enjoyable to experiment with, especially if you're using varied images in your project.
That's all for this week's roundup. We'll include all the links in the show notes. Thank you for spending your time with us today. We'll be back soon. Until then, take care, everyone.
Hosted on Acast. See acast.com/privacy for more information.
* Boston Dynamics and ChatGPT Integration: Boston Dynamics has integrated ChatGPT into its robots. There's a two-minute video showing their team interacting with a ChatGPT-integrated robot that is definitely worth watching.
* ChatGPT Updates: ChatGPT has released new settings. You can now export your data and turn off data collection in the settings. There are also some visual changes they've made. Furthermore, updates have been rolled out to their plugins, including the browsing plugin that now works with ChatGPT-4.
* AI Learning Resources:
* OpenAI and Deep Learning Prompt Engineering Course: This course provides insights into prompt engineering. It's a quick study and is recommended even for those not intending to write code to interact with the API.
* Microsoft's AI for Beginners: This robust document can be found on Microsoft's GitHub page. It covers topics such as neural networks, computer vision, and natural language processing. It is suitable for those who want to delve into the technical aspects of AI.
Links:
Boston Dynamics ChatGPT
OpenAI and Deep Learning Prompt Engineering Course
Microsoft's AI for Beginners
Hosted on Acast. See acast.com/privacy for more information.
In this episode, we dive into some fascinating discoveries, intriguing discussions, and thought-provoking dilemmas surrounding the world of artificial intelligence and the digital future. We start by discussing the latest substantial update to Google's Bard AI, which now boasts impressive coding capabilities and seamless integration with Google Collab and Google Sheets.
Next, we contemplate an insightful tweet by Marie Haynes, an SEO expert, that speculates the possibility of a radical shift in how businesses are found online due to Google's new AI search engine. We ponder the implications of AI interacting with businesses on our behalf and how that could change our experiences.
Finally, we recommend a thought-provoking talk from the Center for Humane Technology by Tristan Harris, who explores the potential catastrophic risks that AI poses to society, the race to deploy AI without adequate safety measures, and the need to upgrade our institutions for a post-AI world. We emphasize the importance of examining the pros and cons of AI and exploring different perspectives on this transformative technology.
Links
https://bard.google.com/updates
https://twitter.com/Marie_Haynes/status/1649755997466947586
Hosted on Acast. See acast.com/privacy for more information.
Hosted on Acast. See acast.com/privacy for more information.
From the publisher's feed
Explore the latest in tech each week with host John Siwicki on the Stack Snacks podcast. This snack-sized audio experience makes cutting-edge AI, product design, and development digestible. Get a…
https://www.youtube.com/@stacksnacks
www.stack-snacks.com
Hosted on Acast. See acast.com/privacy for more information.