What’s on my mind this Sunday?

I'm a child of the cassette era. Well, not a child anymore, but the spirit lives on within me. You remember cassette decks, right? They had that nifty fast forward button to breeze through songs or even skip one entirely. It was a love-hate relationship for me. Sure, I was eager to get to the next fantastic tune, but I couldn't shake the nagging feeling that I was disrupting the sacred "mixtape magic" by not giving each song a fair chance. Maybe the skipped track had something valuable to offer, a lesson I'd never learn.

Now, fast forward to the present, where Godmode and AutoGPT are the newest sensations in the breakneck-paced GenAI world. As we sprint towards the future, I can't help but wonder what we're missing. What are the long-term consequences of GenAI on, say, college and school exams and assessments? In our rush to embrace the new, are we giving up something essential, just as I did with those mixtape gems? Or are we getting to the next best thing?

Ok, time to dig in

A special Ciao 👋 to my NEW SUBSCRIBERS! 
To read previous posts, go here.

Today we will:

1. get into GodMode aka AutoGPT

2. read what students at Cardiff admit about using ChatGPT

3. examine how ChatGPT might impact college exams

4. consider how college assessments shape your future workforce

5. learn how to get GPT-4 for free

6. take psychedelic drugs

GodMode

AutoGPT: Silicon Valley's Latest Obsession Explained

AutoGPT is an experimental application that uses GPT-4 to act autonomously given a goal in natural language. It can perform a task with little human intervention and can self-prompt. While GodMode is an actual product name that offer AutoGPT, many have used it as a synonym for AutoGPT itself.

Silicon Valley’s quest to automate everything is unceasing, which explains its latest obsession: AutoGPT. This open-source app uses OpenAI’s text-generating models, mainly GPT-3.5 and GPT-4, to act “autonomously.”

AutoGPT handles follow-ups to an initial prompt of OpenAI’s models, both asking and answering them until a task is complete. Users provide their goals, and the bot uses GPT-3.5 and GPT-4 to carry out every step needed to achieve the goal. Its capability lies in interacting with apps, software, and services both online and local, like web browsers and word processors.

AutoGPT can even be hooked up to speech synthesizers, allowing it to place phone calls. However, using Auto-GPT requires setup, installation in a development environment, and a paid OpenAI account for an API key.

Although powerful, AutoGPT has limitations and risks. It can behave unexpectedly depending on the objective provided and is prone to inaccuracies due to its reliance on language models. Furthermore, it often doesn't recall how to perform tasks for later use and struggles with breaking down complex tasks.

To demonstrate Auto-GPT in action, let’s show you a comparison between a typical prompt in GPT-4 and one using AutoGPT.

GPT-4 versus AutoGPT

To give you a clear sense of the difference between a typical GPT and AutoGPT, here is a possible breakdown and comparison for how each would interpret and respond to the following prompt: “Please create an entire and complete Christmas meal plan.”

GPT-4:

GPT-4 uses byte pair encoding (BPE) to tokenize text, which means it splits words into smaller units based on their frequency in the text. For example, the word “Christmas” might be split into “Chris” and “tmas”.

To interpret the prompt, GPT-4 would first tokenize it into BPE tokens, such as “Please”, “create”, “an”, “entire”, “and”, “complete”, “Chris”, “tmas”, “meal”, “plan”, “.”.

Then, it would use its neural network to generate a sequence of tokens that is most likely to follow the input tokens, based on its training data and parameters. It would also use its general knowledge and “common sense” to produce a coherent and relevant response.

To respond to the prompt, GPT-4 might generate something like this:

A possible Christmas meal plan is:

• Appetizer: Cheese and crackers, roasted chestnuts, cranberry sauce

• Main course: Roast turkey with stuffing, gravy, mashed potatoes, green beans, carrots, Brussels sprouts

• Dessert: Pumpkin pie, apple pie, whipped cream, eggnog

• Drinks: Mulled wine, cider, hot chocolate

AutoGPT:

To interpret the prompt, AutoGPT would do the same as GPT-4, but it would also break it down into sub-tasks and use the internet and other tools in an automatic loop.

For example, it might search for recipes online, compare different options, check for availability of ingredients, and optimize for cost and nutrition.

To respond to the prompt, AutoGPT might generate something like this:

I have created an entire and complete Christmas meal plan for you. Here are the details:

• Appetizer: Cheese ball with crackers, roasted chestnuts,
  cranberry sauce

• Main course: Roast turkey with stuffing, gravy, mashed potatoes, green beans, carrots, Brussels sprouts

• Dessert: Pumpkin pie, apple pie, whipped cream, eggnog

• Drinks: Mulled wine, cider, hot chocolate

You can find these at your local grocery stores . . (it will then list nearby stores and/or suggest online ordering).

The Expense of AutoGPT

While it sounds like a dream come true, it turns out that Auto-GPT has a few hiccups. For starters, it costs a pretty penny to get things done, even small changes. It also tends to mix up its homework and classwork, wasting time and resources. Plus, it has a bad habit of getting stuck in loops, like a broken record. And don't even get me started on its obsession with python error testing!

It's not uncommon for AutoGPT to take an average of 50 steps to complete a small task. So, the cost of completing a single task is significant. Using baseline costs as shared by OpenAI on how much they charge per 1000 tokens, one AutoGPT session could easily cost $20.This may be a bargain or a waste. Overtime, my bet is that costs per token will decrease while the effort of prompting and responding will get more efficient so we should be able to find harmony - but we are far from that place right now.

Future of AutoGPT

While AutoGPT has its challenges and is not the perfect solution for every application, its potential is undeniable. As AI technology continues to advance, the cost per token and overall efficiency is expected to improve, making it more accessible and practical for a wider range of industries and professionals.

In short, AutoGPT's teamwork skills, like talking to its AI buddies, show promise for future AI success.

Top Takeaways:

• AutoGPT wants to be your AI superhero, but it's not quite ready to save the day.

• It's a bit of a spendthrift, costing more than it should for even the tiniest tasks.

• It mixes up what's due when, jumbling development and production like socks in a dryer.

• AutoGPT can get stuck in loops, like a dog chasing its tail.

• It's a tad too obsessed with error checking, which makes it a bit of a memory hog.

• On the bright side, AutoGPT's chatty teamwork skills with other AI agents could be the key to its future glow-up.

News Vignette 

Moving a news item up today to help intro the following section.

This eye-opening exposé delves into the world of Cardiff University students who have confessed to using ChatGPT to assist them in writing essays. With the help of the AI chatbot, students claim to have achieved first class grades, prompting the university to review its policies and issue new guidance.

As the use of AI in academia raises concerns about academic integrity, the question remains: how will universities adapt to the ever-evolving landscape of artificial intelligence?

Your Future Workforce 
Assessment Vintage

You dear reader, are likely an employer or someone who influences who gets hired in your team and organization. As you no doubt are seeking top talent and future employees in the age of AI, you must remain vigilant and informed about the challenges that AI chatbots like ChatGPT present to education and assessment authenticity.

As we usher in the future workforce, we should ask ourselves:

1. How can we ensure that the talent we hire possesses the genuine skills and knowledge we require?

2. Are we prepared to adapt our recruitment strategies to account for the evolving landscape of educational assessments?

3. And, how can we collaborate with educational institutions to create an environment where assessments remain engaging and relevant, even as AI-generated content becomes increasingly prevalent?

To intro this topic, today, we're venturing into the hallowed halls of education to explore a pressing question for the AI age: how can we keep assessments authentic and engaging in the face of increasingly sophisticated AI chatbots like ChatGPT? 

In two posts below, I’ll first explore the "Exam Apocalypse" and the merits of oral exams, which have both ancient Greek coolness and the ability to foil chatbot cheats. Then, I'll take a detour down the "Write Way" and delve into a smorgasbord of alternative assessments, showcasing a diverse range of methods that aim to keep students on their toes and AI at bay.

👩‍🏫 Exam Apocalypse: Why Oral Exams Might Save Us from Chatbot Cheating

Picture this: you're a student, stressed out by the dreaded finals season, only to realize that you won't be writing any essays this time around. Instead, you'll face a panel of examiners who have actually read your work and will ask you questions about it. That's right, we're talking about oral exams, the underdog of assessment methods.

Let's dive into a thought-provoking review of reasons why we should rethink exams and embrace oral assessments:

🏛️Ancient Greek Coolness Factor: Oral exams, or viva voce, date back to ancient Greece, where philosophers defended their knowledge in public oral defenses. That's some 2,000-year-old swagger right there. Who wouldn't want to follow in the footsteps of Socrates and Plato?

🗣️Talk the Talk: Written exams can be a drag, and students often struggle to express their thoughts on paper. With oral exams, students can showcase their knowledge through conversation, which is way more engaging than scribbling essays with aching hands.

🤖AI-Proof: As AI chatbots like ChatGPT become more advanced, it's getting harder to prevent students from using them to cheat on written assessments. Oral exams, on the other hand, require students to answer questions on the spot, making it impossible for ChatGPT to swoop in and save the day.

🚫Bye-bye, Cheating: In studies on oral exams, instances of cheating were virtually nonexistent. Say goodbye to sneaky crib sheets, writing-on-forearms tactics, and other creative cheating methods that would make MacGyver proud.

📝Reflection Time: Oral exams offer the chance for students to create and share reflective journals, which can't be replicated by chatbots. This helps ensure that the work is genuinely their own, and they can't just regurgitate someone else's thoughts.

⏱️Speedy Grading: Examiners can assess oral exams relatively quickly compared to written exams, which means less time buried under a mountain of papers to mark. Plus, the weekend is saved for binge-watching your favorite Netflix series or taking that much-needed nap.

🎢Lasting Impact: Though nerve-racking, students tend to remember oral exams more vividly, making the knowledge stick better. It's like a rollercoaster ride – terrifying in the moment but memorable and exhilarating in the end.

Oral exams offer an exciting, AI-resistant alternative to traditional written assessments. As students face the challenges of ChatGPT and other AI-generated essays, let's bring back the ancient art of conversation and make assessment engaging and authentic again. And who knows, maybe oral exams will even make students feel a bit like ancient Greek philosophers, minus the togas.

The Write Way? Debunking the Oral Exam Hype and Exploring a World of Alternatives

Sure, we've waxed poetic about oral exams and their ancient Greek coolness factor, but let's take a step back and explore the broader world of alternative assessments. After all, the educational landscape is vast, and there are plenty of other ways to evaluate students' skills and knowledge.

Here's the scoop on the pros and cons of various assessment methods, as well as some tips for improvement:

📚 Coursework & Extended Essays: These projects showcase students' research and writing skills and give them a taste of real-world settings. But beware, procrastinators may suffer, and the potential for plagiarism lurks around every corner.

🗣️ Oral Exams: As we've mentioned, these conversation-based assessments can be engaging and AI-resistant. However, they may be more prone to examiner bias and cause sweaty palms and shaky knees for the more introverted among us.

🎨 Portfolios: A collection of students' best work highlights their progress and skills over time. But, assembling a portfolio can be time-consuming, and the grading process might be more subjective than traditional exams.

🎭 Performance-Based Assessments: These tasks emphasize problem-solving and teamwork in real-world situations. But, practicality can be an issue, as not all subjects lend themselves to performance-based assessments (good luck performing a calculus equation!).

🔧 Skill Development Assessments: Focusing on developing practical skills relevant to students' future careers is great, but it's not always easy to measure these skills consistently across different learners.

Returning to the opening questions, addressing these critical questions is essential for securing a skilled and reliable workforce.

By acknowledging the potential impact of alternative assessments and supporting innovative solutions, we can contribute to the development of a credible educational system that prepares students to become the accomplished professionals our organizations need in the AI-driven future.

I urge you to consider what role you might play or what influence you might have on how our education system reacts to the impact of AI. Don’t take it for granted that things will magically happen - this is a massive development that will take educators, employers, and students working together to forge a better way.

News you can Use: 

Access GPT-4 for Free

Forefront AI Launches Free Access to GPT-4 with their New Forefront Chat Platform

New York-based software development company, Forefront AI, recently launched Forefront Chat, providing users with free access to GPT-4. The platform also offers image generation, custom personas, and shareable chats. By logging on to chat.forefront.ai and creating an account, users can experience the powerful AI capabilities of GPT-4 without any subscription fees.

News you can Lose: 

Silicon Valley’s Other Obsession: Psychedelic Drugs

Sam Altman, chief deity of OpenAI, ascended to the throne of Journey Colab, partnering with a luxury rehab clinic to guide MDMA and psilocybin through trials and create a model for administering psychedelic drugs to mere mortals. This was part of a larger trend in Silicon Valley, where venture capitalists bet millions on the transformative power of these mystical therapies.

That's a wrap for today. Stay thirsty & see ya next time! If you want more, be sure to follow me on Twitter and LinkedIn.

DISCLAIMER: None of this is legal advice. This newsletter is strictly educational and is not legal advice or a solicitation to buy or sell any assets or to make any legal decisions. Please /be careful and do your own research7