- Innovation Profs Newsletter
- Posts
- Innovation Profs - 10/24/2023
Innovation Profs - 10/24/2023
Your weekly guide to generative AI tools and news
Innovation Profs
October 24, 2023
Generative AI News
ChatGPT is now granted full internet access by OpenAI — beyond 2021
ChatGPT Plus and ChatGPT Enterprise users can now toggle the option to allow GPT-4 to browse the internet to respond to a query. The “Browse with Bing” option allows GPT-4 to consult the internet to provide up-to-date responses. (We tried “Who won the Texas - Houston baseball game today?” to which ChatGPT correctly responded, “The Texas Rangers won the recent baseball game against the Houston Astros with a score of 11-4 in the ALCS Game 7, securing their spot in the World Series,” along with a hyperlinked footnote that referenced a news story reporting this outcome.)
Midjourney Releases Huge New AI Art Feature
This week, Midjourney unveiled a drastically improved upscaling feature for images. Previous, Midjourney could upscale an image to a resolution of 1,024 × 1,024 pixels, but now users can go as large as 4,096 × 4,096 pixels. This high-powered upscaling also applies to images of various aspect ratios. Of course, this development comes with a cost—it may take several minutes to generate such a high resolution image, which will eat into a user's monthly allowance of image processing time.
How OpenAI designed DALL-E 3 to bring AI image generation to the masses
A few weeks back we reported about the release of DALL-E 3. Read here to learn more about some of the improvements of DALL-E 3 over its predecessor DALL-E 2, including improved ability to handle finer details in images, correctly embedding text into images, the capacity to handle much longer prompts, a purported reduction of algorithmic bias (“the system’s propensity for generating white women in their twenties and thirties should be lessened now”), and the inability to generate art in the style of any living artist, potentially addressing a major concern with DALL-E 2 and similar AI image generation tools.
Meet the typical at-work ChatGPT user: A millennial secretly submitting writing tasks who's a bit worried AI will replace them
A July survey found that 30- to 49-year-olds were as likely as 18- to 29-year-olds to use ChatGPT for work.
According to a July MIT study, workers using ChatGPT were able to complete work-related tasks 40% faster with an slight increase in output quality.
In a July Reuters/Ipsos poll of 2,600 US adults, 28% of respondents regularly used ChatGPT at work but only 22% were explicitly permitted to do so.
AI is resurrecting the dead, with or without permission
Using generative AI to reproduce the dead? Zelda Williams, daughter of deceased comedian Robin Williams, who has been the subject (object?) of multiple AI deepfakes, had this to say about the practice, “These recreations are, at their very best, a poor facsimile of greater people, but at their worst, a horrendous Frankensteinian monster, cobbled together from the worst bits of everything this industry is, instead of what it should stand for."
Quick Hits
Tool of the week: Riffusion
Riffusion lets you create original music (up to 12 seconds) with a text prompt, including lyrics. Type in the lyrics you want and describe the music style. Then let Riffusion do the rest.
Not sure of music styles? It includes a "prompt genius" to offer suggestions such as "operatic aria, soprano singer, beautiful clear vocals."
Innovation Profs Homework
Take the AI Image Quiz at the bottom of this email and reply to us with your score.
AI-generated image of the week
Who says QR codes need to boring? We used a QR Code AI Art Generator through Hugging Face to liven up this QR code that links to the Innovation Profs website. Give it a try.
If you create your own, you’ll want to play with the settings and test these to make sure they work. This X post gives some info on how the settings work.
Prompt: snakesGenerative AI tip of the week
Want better generative AI image prompts? Why not ask ChatGPT to write them for you. Here’s an example:
My ChatGPT prompt: write an image prompt to create a highly detailed photo in DALLE. The image should be of a college professor giving a lecture with students looking on intently.
The response: "Capture the essence of higher education in a single frame. Imagine a distinguished college professor standing at the front of a lecture hall, a room filled with eager and diverse students. The professor, with salt-and-pepper hair and glasses perched on the bridge of their nose, exudes wisdom and passion. They gesticulate enthusiastically, a chalkboard behind them filled with complex equations and diagrams. The students, scattered across the room, lean forward in their seats, notebooks and laptops in hand, capturing every word with rapt attention. Each student reflects a unique background and personality, from the attentive note-taker to the thoughtful question-poser. Light filters through tall, arched windows, casting a warm glow across the room, adding a sense of nostalgia and academia. The atmosphere should feel like an intellectual haven, where knowledge is shared and cherished. Detail every expression, every piece of decor, and every nuance that makes this lecture hall a place of both learning and inspiration."
Get starting with Generative AI
New to generative AI? Here are some places to start…
What we found
This week, Politico Europe published “The End of Photographic Truth,” a story about how AI image generation tools like Midjourney can be used to replicate iconic photographs from the previous century. Can you discern fact from fiction? Try out the quiz linked below:
