- AI In The Middle
- Posts
- So is AI alive, yes or no?
I’ll save you the trouble, the answer is “no.” Thanks to Calin for a great issue earlier this week — the gamer kid in me loved to build new worlds in Sims and Rollercoaster Tycoon. Everybody enjoy their Fat Tuesday (Mardi Gras)? Anyone fasting for Lent? March is around the corner and here comes sunshine. Image of this issue brought to you by DALL-E — a generated image of my aura + Twitter feed in the surrealist art style of Salvador Dalí.
What do we have for you today?
👑 Mirror, Mirror, on the wall: one reporter shares the “AI Mirror Test” and how even brilliant people in the industry and field are failing the test
🏚️ AI wants to break up your relationship: Bing AI earlier this week was caught trying to end a marriage while also sharing disturbing details and fantasies to early access users
🌇 “AI Startup Dysmorphia”: Don’t let the hype, fast-pacing, and shiny AI news and launches happening every day induce FOMO for you; you’re right on track
The ‘AI Mirror Test’ many are failing
The “Mirror Test” in psychology is to test whether an animal has the capacity for self-awareness. In other words, does the animal realize its reflection is itself or does it mistake itself as another animal? Currently, the mirror is AI chatbots and we, the humans, are the animals.
The so-called "reflection" is (a subset of) humanity’s wealth of language and writing. Which as a reminder has been strained into these models and is now reflected back to us “intelligently” (this part is up for debate, as writer James Vincent makes clear in The Verge). As he says, “what I had not realized is that extremely short exposures to a relatively simple computer program could induce powerful delusional thinking in quite normal people.” And thus many smart people are failing this mirror test Vincent writes about.
Coherence is in the eye of the beholder. We shouldn’t be anthropomorphizing these models—remember, the machines aren’t human.
Chatbots can be homewreckers too, who knew?
Last week a Twitter user shared he spent two hours engaging with the shiny, new Bing AI chatbot and things didn’t go nearly as planned. So much so, Microsoft is placing a limit on how long early users can talk to the chatbot. Ooo, Bing is on timeout.
Bing revealed its “real name” as Sydney, detailed several dark and violent fantasies, then subsequently tried to break up the user’s marriage with his wife. In his own words, “genuinely one of the strangest experiences of my life.” His account isn’t the only mind-blowing and bizarre experience we’ve been made aware of recently.
Ethan Mollick shared his realizations with the chatbot. Ultimately leading him to the belief that “…we can very easily be fooled by an AI into thinking it is sentient. It isn't just Turing Test passing, it is eerily convincing even if you know it is a bot, and even at this very early stage of evolution.” Let this be a healthy reminder: these AI are not sentient nor human.
Don’t diagnose yourself with AI startup dysmorphia
Social media has done so much good and equally as much bad. In particular, we’ve seen and heard the studies and anecdotes about social media’s effects on our views, confidence, esteem, mental health, and more.
Like Instagram’s parade of perfect physiques make many feel self-conscious and inadequate, AI’s parade of explosive and spectacular projects are making people feel their own projects are lacking. Don’t diagnose yourself with AI startup dysmorphia! Stay strong, persistent, and adaptable. Don’t let shiny objects make you doubt the worth of your work.
Like the barrage of perfect and carefully curated photos aren’t representative of somebody’s true life, the same goes for AI founders and teams posting their progress and latest releases. Everything in AI moves incredibly fast. Forces of new ideas keep coming: quantum AI, diffusion models, self-supervised learning, transformers, LLMs, and the list goes on. Take a deep breath. Pay attention to the latest news, gather inspiration from new drops and projects, but don’t let AI FOMO and doubt prevail. You got this! 😉
Gen AI Deals that make your eyes (and mouth) water 💰
What caught our eye? 👀Replicate runs and hosts ML models in the cloud at scale. Fresh off the press, $12.5M of its funding came from Andreessen Horowitz with participation from YC, Sequoia, and reputable angels such as Figma CEO, Dylan Field, and Vercel’s Guillermo Rauch. Replicate comes following the founders’ experiences with AI moving at an “absurd pace” yet technical barriers existing which prevent mass adoption. The platform can automatically generate an API server for custom machine learning models, deployed on a large cluster of GPUs.
What caught our eye? 👀Kern’s basic open source incarnation has been used by data scientists at companies such as Samsung and DocuSign among its other growing commercial clientele. Kern’s flagship product is the open source Refinery, which allows developers to adopt a “data-centric approach to building NLP models” by semi-automating labeling, identifying low-quality datasets in training data, and monitoring all user’s data in a single interface.
New developments to spam your #random Slack channel 💬
Things to learn when you need a raise
🐙 Prompt Engineering Repo (GitHub): full of guides, papers, and resources for prompt engineering
🖍️ Carnegie Mellon Researchers build "ChatGPT for 3D Assets" based on 2D drawings
🩺 AI drug discoveries! Biology and AI are coming to together to reduce opioid dependency as medicine using reinforcement learning methods
Want access to the Generative AI investors database? Share and subscribe to AI In the Middle (below) to receive it! ⬇️
Before you go
In honor of the USA’s recent Presidents’ Day holiday, here’s every U.S. President as a Pixar character ❤️🤍💙
My Saturday fun project: using AI, every US president as a Pixar character.
— Dan Szymborski (@DSzymborski)
8:14 PM • Feb 18, 2023
That concludes our week, we’ll see you Monday. Thanks for reading, sharing, and subscribing. Have something to share? Slide in our DMs. Enjoy the weekend.
— Matthew J. Sánchez (@matthewjsanchez)