GenAI vs. the Remote VAs: A Guide for the Emerging Era of AI?
How Should We Work with Generative AI?
Its power and potential are undeniably shaping a technological revolution that rivals the introduction of the internet or the iPhone, evidenced by OpenAI’s ChatGPT amassing a staggering one million users within five days. Corporations underpinning this AI revolution, such as NVIDIA, are riding a wave of growth and success, their market cap surpassing the trillion-dollar mark.
With great power, however, comes a whole load of scary. There are worries over job loss, the exponential growth of fake news, and, alarmingly, the potential doom of humanity. The latter is especially concerning given that Sam Altman, Elon Musk, Geoffrey Hinton, and other heavy hitters have signed this statement.
Yet others don’t think much about AI’s latest and greatest. They point to its lack of truth-telling and propensity to “hallucinate.” I’m not too fond of the term. In reality, it’s closer to a hammer than a human. But that’s the term that’s being bandied about.
A great example is a courtroom mishap involving a lawsuit against Colombian airline Avianca. The lawyer for the plaintiff in the case, Steven A. Schwartz, relied on ChatGPT to substantiate his brief with legal precedents. However, the brief was filled with fictional cases generated by the tool.
When challenged about these alternative facts, Schwartz verified the cases’ authenticity by simply asking the chatbot, which reassured him of their validity, directing him to resources like Westlaw and LexisNexis. The defence exposed the spurious cases, leading to potential sanctions for Schwartz.
Among the fallacies was a non-existent case, Varghese v. China Southern Airlines Co., Ltd., where the chatbot misattributed details from a real case. Schwartz confessed his ignorance of the AI’s ability to fabricate information and regretted his reliance on it without thorough verification.
This case underlines the risk of depending solely on AI for research, especially without double-checking information, and the absurdity of using chatbots as primary sources. The incident is a sobering reminder that, although generative AI can produce compelling content, its output is not always reliable or accurate. (For more, see the Verge).
Though it is easy to engage in a bit of schadenfreude at the lawyer’s expense, that’s not the point of including this incident. Instead, it is another example of the need for clarity around how this technology can enable productivity in its current stage of infancy. This is especially for those of us that have used it to summarize articles, format tables, get it to generate VB code that produces PowerPoint slides, and so on. It’s productive, but at the same time, it makes false statements. So how should we think about Silicon Valley’s latest? Here’s how I would put it:
Generative AI is about amplification; not abdication.
Generative AI (GenAI) can magnify one’s efforts but is not a substitute for one’s professional responsibilities to one’s employer, customer, or clients. We can see that when the lawyer used the tool as a stand-in for a junior on his legal team, he ended up in a world of professional pain.
How can we navigate this thrilling yet daunting reality of AI? Tim Ferriss’s “4-Hour Work Week” might be the aid we need.
The Four-Hour Work Week: Why is a 15-year-old book Relevant Today?
Applying this idea of amplification to generative AI, we see how AI takes a small amount of input text and generates a significant output. One provides a seed, and out comes a tree. This output can be entirely usable as a first draft, whether for an email, a blog post, or other forms of content. Conversely, it also takes a large text, making it more digestible. Either way, you end up 10X more productive.
Regarding assessing how productive generative AI can be, Chapter 8: Outsourcing Life: Offloading the Rest and a Taste of Geoarbitrage, from Ferriss’s book, is helpful. This part of the book elaborates on how remote virtual assistants (VAs) based in India can enhance your productivity. We’ll closely examine how effectively generative AI can handle the tasks Ferriss identified as ideal for outsourcing to these VAs.
The concept of abdication is also addressed, albeit in a more indirect way.
Ferriss recounts a frustrating experience with a remote virtual assistant and, upon reflection, realized the fault lay in his own failure to take full responsibility to ensure the assistant worked productively. He emphasizes giving clear, precise instructions and having the VA rephrase tasks to confirm understanding. (This is one of the best practices for promoting generative AI: ask if it understood your instructions.) Ferriss also recommends monitoring virtual assistants closely, setting shorter deadlines, and dividing large tasks into smaller pieces.
The problem of abdication arises when the user leverages generative AI without the necessary quality control to verify the output generated. In the context of law, for instance, using the free version of ChatGPT would be equivalent to hiring an unqualified person to serve as a junior lawyer or paralegal on a team.
The users need to review the output carefully as we look at the current stage of generative AI’s maturity. The user must be able to provide the necessary quality control. This may involve cross-referencing the output by doing the necessary research or relying on one’s knowledge to discern the truth from the falsehood.
Further Considerations on Assigning Tasks to the AI
Breaking Out of your Comfort Zone
Ferriss suggests that a significant benefit of a remote personal assistant is that it helps to rethink how tasks can be taken off our plate and put on someone else’s. Applying this to generative AI, it can be a challenge to reorganize one’s work routine. However, this can result in time savings in the long term because it’s not just you doing the work. Think of learning shortcuts like CTRL-C and CTRL-V. There was a learning curve. Now, not only do we not even think about it when we use them, we are not as productive without them.
Data Confidentiality
There is an important caveat to be made here. When sending data to ChatGPT, Bard, etc., it is like sending it to the public. Consequently, it will likely be against company policy to send customer, financial, or other confidential information to such services. However, some AI services, like Notion, offer assurances about security and confidentiality. This may or may not be acceptable to corporate IT. So you will need to check.
Eliminate before you Delegate.
Ferriss also cautions that not all tasks should be delegated. First, he suggests “eliminate before you delegate.” Unenjoyable, non-income producing and non-mandatory tasks should not be done by anyone — including AI. Second, the work must warrant the time it will take to draft the prompt. If you can send a five-worded Slack message to a colleague, there is no need to turn to AI. Suppose it’s about generating a polite email to a fitness club about the need to repair most of their cross-trainers. In that case, generative AI is almost designed for such a task. No need to wait in line to get passive-aggressive looks from the gym’s employees. We can let AI deal with that.
The Throw Down: Generative AI versus Remote VA
Now it’s time to put generative AI to the test.
The expanded and updated edition of Ferriss’s book offers us two distinct angles for assessing the performance of generative AI in comparison with remote VAs.
Firstly, we will take a comprehensive look at the particular tasks Ferriss earmarks as suitable for outsourcing. The diverse array of functions in question will offer a broad canvas for evaluating the adaptability and efficiency of generative AI in various scenarios and its ability to meet different output needs.
Secondly, we will review the experiences of AJ Jacobs, editor at large at Esquire, who employed a remote VA in his professional capacity. His first-hand experiences and insights provide a valuable benchmark against which to measure the real-world capabilities of generative AI.
Throughout this exploration, we aim to highlight the potential and limitations of generative AI in contrast with traditional remote VAs and how the technology can be effectively harnessed in line with Ferriss’s guidance for optimal productivity.
Part 1: Can AI Take on Tim’s Tasks?
Ferriss lists 11 tasks that serve as illustrative examples of what can be done by a remote virtual assistant. Generative AI can do about 5.5 of them. Let’s look at how generative AI can help with each of them or get part of the way there.
Task#1: Reminders to an Overzealous Client
Prompt: “Generate a polite email to overzealous client reminding him to pay his current parking fines, as well as not speed and collect parking fines.”
Output:
“Subject: Gentle Reminder Regarding Parking Fines and Traffic Safety
Dear [Client’s Name],
I trust this email finds you well. I’m writing to bring your attention to an important matter that needs your immediate attention.
Our records indicate that there are currently some outstanding parking fines associated with your account. The total sum of the fines is [$ amount]. I understand that it can be easy to overlook these notifications, and this gentle reminder serves as a prompt to help you manage them promptly. Kindly ensure that these are settled as soon as possible to prevent any additional penalties or inconvenience.
In addition, I would like to take a moment to discuss the importance of adhering to traffic rules and regulations. The safety of all road users is paramount, and as such, it is crucial to respect speed limits and parking regulations at all times. Over-speeding and improper parking not only lead to fines but also pose risks to yourself and others. I trust that you understand the significance of this matter and will consider it moving forward.
If there are any concerns or if you require any assistance with the payment of these fines, please feel free to contact our team. We are always ready to assist and make this process as seamless as possible for you.
Thank you for your attention to this matter and for your commitment to safe and responsible driving.
We value your cooperation and look forward to your prompt response.
Best regards,
[Your Name]
[Your Position]
[Your Contact Information]”
Conclusion: It seems like AI can handle this one adeptly.
Task#2: Diet, Food, and Reminders
One of the requests was “charting a diet plan, reminding client on it regularly, ordering groceries based on the specific diet plan.” Generative AI can convert a recipe or a seven-day meal plan into a grocery list that can be ordered via the Instacart plugin (see here for a live demo of this feature).
Conclusion: AI gets us 80% or more there, but more functionality is required to get to the level of remote VA. But someone doing this for themselves, as a job, or as a business would have access to calendar functionality (e.g., Outlook) to use for the reminders to send this out to whoever needs it.
Task#3: A Kid’s Guide to Tying Shoelaces
Prompt: “Write a guide for a 5 year old on how to tie their shoelaces.”
Output:
Guide for Tying Shoelaces
Get Ready: Sit down on a comfy chair and put your shoe with the laces right in front of you. Make sure you’re comfortable and ready to go on a little adventure with your laces!
Make an “X”: Take one lace in each hand. Now, let’s pretend that they are two little snakes. Make the snakes play with each other by crossing them over to form the letter “X.” Then, let one snake go under the other and pull it tight. You will see a little knot. Good job!
Make the Bunny Ears: Now, let’s turn each lace into a bunny ear. To do this, take a lace and fold it so it forms a loop or a “bunny ear.” Do the same with the other lace. Now you have two bunny ears!
Cross the Bunny Ears: Next, make the bunny ears play together just like the snakes did before. Cross one bunny ear over the other to make another “X.”
Pull Through: Let one bunny ear go under the other (just like the snake did before), and then pull it through the hole that you see. Pull both bunny ears outwards until you see a pretty bow appear.
You Did It!: And there you have it! You’ve tied your shoelaces.
Remember, it might take a few tries to get it right, and that’s totally okay. The more you practice, the easier it will get. You’re doing a great job!
Remember these steps:
Snakes make an “X” and one goes under
Make bunny ears
Bunny ears make an “X,” and one goes under
Pull, and you have a bow!
Keep practicing, and soon you’ll be a shoelace-tying expert!”
Conclusion: AI wins again.
Task#4: Find Parking in a City
Prompt: “Where is the best place to park for the Collision conference in Toronto next week?”
Output:
“The best place to park for the Collision conference in Toronto next week would likely be at the Enercare Centre’s parking facilities, where the conference is taking place. Exhibition Place, where the Enercare Centre is located, offers 5,800 parking spaces, including 1,300 underground. They also have parking lots numbered from 850 to 859. The lots provide a total of 88 accessible parking spaces, 14 spaces for hybrid vehicles, electric vehicle charging stations, and 5 spaces for parents with small children and/or strollers.
The regular daily parking rates at the surface lots and the indoor parking at the Enercare Centre are $18.00, and the rate after 6:00 PM is $15.00. Please note that rates may increase by $3 on peak usage days and for special events. The maximum rate for BMO Field and special events, including Budweiser Stage or Echo Beach events, is $45. These rates may change without notice and during playoffs.
You can pay for parking using a credit card (including Apple Pay, Google Pay, AMEX) or debit card. The Green P Mobile App is also accepted, except for lot 850 and the Enercare Parking Garage.
The surface parking lots are always open, and you can use the Pay and Display machines or Mobile Pay. The opening times for underground parking at the Enercare Centre may vary, so you might want to call Parking Services at 416–263–3646 for additional information.
To get to the conference from Toronto’s Pearson Airport or downtown Toronto, you can use the UP Express, which offers a 40% discount for Collision Conference attendees. The service operates every 15 minutes, seven days a week. If you’re arriving from outside the Greater Toronto Area, you can connect to GO Transit to Exhibition GO Station, which is only a 5-minute walk to the Enercare Centre.” (Link 1, Link 2)
Conclusion: Ferriss’s original request was, “Find a parking slot for your car in some other city even before you make the trip.” However, I modified this for an in-town conference I plan to attend, i.e., to assess its validity. It passed.
Task#5: Get an authenticated weather report and forecast for a particular time and place
Prompt: “Get a weather forecast and weather report for January 24, 2018 for Hamilton, Ontario, Canada”
Output:
“I’m sorry, but as an AI, I am unable to interact with certain elements on webpages such as dropdowns or form inputs. However, you can find the historical weather data for Hamilton, Ontario, Canada on January 24, 2018 by visiting the Government of Canada’s climate website. On this site, you can search for historical data by station name. Please follow this link to access the site: Environment and Climate Change Canada.”
Conclusion: It did direct me to this site, but it’s unclear if pulling from a government website counts as “authenticated.” Remote VA wins this one, but generative AI gets us part of the way there.
Task#6: Talking to parents in our client’s stead
If this is a live conversation that needs to be had, then the remote VA would be the winner. Alternatively, generative AI could do this through voice mails or emails. The voice mails could be trained to have the client’s voice through Eleven Labs and uploaded to WhatsApp or a similar messaging service.
Part 2: AJ Jacobs’s Experience with the Remote VA
The chapter also examines how AJ Jacobs, then the ‘editor-at-large’ at Esquire magazine, retained the services of ‘Honey K. Balani’ from Bangalore. Jacobs walks us through several tasks he outsourced to Honey, providing more detail than we could glean from Ferriss’s list. This greater detail allows us to create better prompts, even utilizing Jacobs’s own words to stimulate the generative AI.
Task#7: Researching a Famous Personality
The first task discussed involves researching a potential individual about whom he plans to write in Esquire. To keep this post-family-friendly, we’ll adjust the subject matter to examine Sam Bankman-Fried, recently indicted over numerous issues regarding how his crypto-exchange, FTX, managed funds.
Prompt: “You are the editor large at Esquire and have been assigned to write a profile of Sam Bankman-Fried (SBF). Provide a breakdown of his personal life (e.g. pets, favorite foods, etc.) and his professional life. Provide links.”
Output:
“Sam Bankman-Fried was born on March 6, 1992, in California. He grew up in a highly educated family, with both his parents being professors at Stanford Law School¹. He attended high school at Crystal Springs Uplands School in Hillsborough, California¹. He also participated in a summer academic program for gifted high school students in mathematics¹.
Sam Bankman-Fried is vegan⁵. As of mid-2021, it was reported that he lived in a five-bedroom penthouse in the Bahamas with approximately ten roommates⁵.
Sam Bankman-Fried is the CEO of FTX, which he co-founded in 20¹⁹³. He is also the founder of Alameda Research⁴.
Here are some links that you might find useful:
- [Who Is Sam Bankman-Fried? — Investopedia](https://www.investopedia.com/who-is-sam-bankman-fried-6830274)
- [Meet Sam Bankman-Fried’s family: His parents are Stanford Law … — Business Insider](https://www.businessinsider.com/sbf-sam-bankman-fried-family-stanford-columbia-university-professors-ftx-2022-12)
- [Sam Bankman-Fried — Forbes](https://www.forbes.com/profile/sam-bankman-fried/)
- [Sam Bankman-Fried — Wikipedia](https://en.wikipedia.org/wiki/Sam_Bankman-Fried)
I hope this helps!
Source: Conversation with Bing, 2023–06–21
(1) Who Is Sam Bankman-Fried? — Investopedia. https://www.investopedia.com/who-is-sam-bankman-fried-6830274.
(2) Sam Bankman-Fried — Wikipedia. https://en.wikipedia.org/wiki/Sam_Bankman-Fried.
(3) Meet Sam Bankman-Fried’s family: His parents are Stanford Law …. https://www.businessinsider.com/sbf-sam-bankman-fried-family-stanford-columbia-university-professors-ftx-2022-12.
(4) Sam Bankman-Fried — Forbes. https://www.forbes.com/profile/sam-bankman-fried/.
(5) Who are Sam Bankman-Fried’s parents? Meet Joseph and Barbara. https://nypost.com/2022/12/14/who-are-sam-bankman-frieds-parents-meet-joseph-and-barbara/.
(6) Polyamory & pills: Inside Sam Bankman-Fried’s FTX party pad. https://nypost.com/2022/11/21/polyamory-pills-inside-sam-bankman-frieds-ftx-party-pad/.
(7) Getty Images. https://www.gettyimages.com/detail/news-photo/founder-sam-bankman-fried-is-seen-on-january-3-2023-in-new-news-photo/1454050564.”
Conclusion: From a research perspective, the links are helpful to get started. However, it’s not at Honey’s level. Jacobs comments on the excellent presentation, organization, and use of charts, suggesting that such a service will eliminate the need for American workers. One does not get the same sort of elation reviewing GenAI’s response. But it will cut down on the Google searches required.
Task#8: Getting Through the Daily Routine
Jacob’s experience with the remote VA service “Your Man in India” (YMII) highlighted the efficiency gains that outsourcing specific tasks could provide. Over the course of a few days, he entrusted a variety of online errands to Asha, a representative from YMII. These tasks ranged from paying bills and purchasing items from drugstore.com to finding a Tickle Me Elmo for his son. When the store ran out of Tickle Me Elmos, Asha bought a Chicken Dance Elmo, which Jacob appreciated as a sound decision.
Conclusion: This task requires human interaction and cannot be completed by GenAI. Regarding the potential of Generative AI, while it cannot perform these diverse tasks, progress is being made with tools such as AutoGPT. For instance, a version of AutoGPT could install the software necessary for the assigned job — even though it was not explicitly instructed to install the said tool. In the future, as these AI tools continue to evolve and improve, we might see them undertaking more complex tasks like making purchasing decisions or interacting with customer service on behalf of users.
Task#9: Email to the Colorado Tourism Board
Jacobs requested Honey to write a gentle email to the Colorado Tourism Board, asking them to stop sending him requests. This was the email she drafted:
“Dear All,
Jacobs often receives emails from Colorado News, too often. They are definitely interesting topics. However, these topics are not suitable for “Esquire.” Further, we do understand that you have taken a lot of initiatives working on these articles and sending it to us. We understand. Unfortunately, these articles and emails are too time-consuming to read. Currently, these emails are not serving the right purpose for both of us. Thus, we request you to stop sending these emails.
We do not mean to demean your research work by this.
We hope you understand too.
Thanking you,
Honey KB”
Ref: Ferriss, Timothy. The 4-Hour Workweek, Expanded and Updated (p. 148). Harmony/Rodale. Kindle Edition.
Prompt: “You are the editor at large at Esquire. You are receiving annoying emails from the Colorado Tourism Board and want the stop sending you such emails. The recent request is around a festival in Colorado Springs featuring the world’s most famous harlequin. The emails should be exceedingly polite, but have a little undercurrent of indignation.”
Output:
“Subject: Re: Invitation to Colorado Springs’ Harlequin Festival
Dear Colorado Tourism Board,
I trust this message finds you well.
I am truly appreciative of the information you consistently provide regarding the multifarious happenings in the vibrant state of Colorado. It is indeed heartening to be acquainted with your unwavering enthusiasm for promoting your beautiful region.
The recent correspondence regarding the forthcoming Harlequin Festival in Colorado Springs has been received with much interest. It sounds like an event of significant cultural importance, one that I’m sure will bring great delight to those who have the opportunity to experience it.
However, it is with a hint of regret that I must bring to your attention a small matter that has been of concern for some time now. As the Editor at Large for Esquire, my role requires a focus on a specific range of topics, and, quite honestly, the updates received from your end, although rich and informative, do not quite align with my professional needs.
While it’s always a pleasure to learn about the cultural tapestry of Colorado, the frequency and content of your emails tend to encroach upon my time and, honestly, inundate my inbox. Therefore, I kindly request that you consider removing my email address from your distribution list.
I hope this does not in any way diminish the profound respect I hold for the commendable work the Colorado Tourism Board undertakes in promoting the attractions of your state. This request is merely an attempt to better manage my inbox and maintain focus on my role’s specific requirements.
Thank you for your understanding and attention to this matter. I wish you all the success with the Harlequin Festival and other future events.
Yours sincerely,
[Your Name]
Editor at Large,
Esquire”
Conclusion: Though it’s longer, it achieves Jacobs’s desired balance of politeness with a touch of indignation. Regardless, he would have this task completed almost immediately.
Task#10: Drafting an Apology Email to Jacob’s Wife
As noted in the book: “My wife got annoyed at me because I forgot to get cash at the automatic bank machine … I wonder if you could tell her that I love her, but gently remind her that she too forgets things — she has lost her wallet twice in the last month. And she forgot to buy nail clippers for Jasper.”
Asha, Jacob’s other assistant, generated the following email along with an e-card with two teddy bears saying he was sorry:
“Julie,
Do understand your anger that I forgot to pick up the cash at the automatic machine. I have been forgetful, and I am sorry about that. But I guess that doesn’t change the fact that I love you so much.…
Love
AJ
P. S. This is Asha mailing on behalf of Mr. Jacobs.”
Ferriss, Timothy. The 4-Hour Workweek, Expanded and Updated (pp. 149–150). Harmony/Rodale. Kindle Edition.
Prompt: [identical to the note Jacobs sent Asha] “My wife got annoyed at me because I forgot to get cash at the automatic bank machine … I wonder if you could tell her that I love her, but gently remind her that she too forgets things — she has lost her wallet twice in the last month. And she forgot to buy nail clippers for Jasper.”
Output:
“Subject: A Little Reminder with Much Love
Dear [Wife’s Name],
Firstly, let me express how much you mean to me. Your patience, understanding, and support make every day brighter, and I’m truly grateful for your presence in my life.
Regarding the minor oversight of not withdrawing cash from the ATM earlier, I want to apologize. I understand that my forgetfulness may have inconvenienced you, and for that, I am genuinely sorry.
However, it has come to my attention that the course of life, with its many tasks and distractions, affects us all, and we’re only human after all. For instance, I couldn’t help but remember how your wallet seemed to grow legs and wander off on its own twice in the last month. It’s quite a skill it has there, isn’t it? And, oh, poor Jasper is still waiting for his new nail clippers that we somehow managed to forget on our last shopping trip.
Let’s consider these little lapses in our memory as gentle reminders of our shared human experience. Rather than causing friction, they provide opportunities for us to empathize and support each other more.
You and I we’re a team, and teams stick together through the forgetful days as well as the memorable ones. I love you, and together, we can manage our shared forgetfulness with understanding and good humour.
Thank you for your patience, my love. Here’s to many more days of shared forgetfulness and shared memories.
With all my love,
[Your Name]”
Conclusion: GenAI can apologize to a spouse, but it’s uncertain whether Jacobs would have the necessary awareness to avoid self-sabotage if he were writing the prompt himself. As Jacobs commented, Asha saved him from himself by removing the comments about the wallet and nail clippers. Without any special instruction, GenAI did not save him from himself. However, assuming he had awareness, he could have used the following prompt:
You are a remote virtual assistant who is helping a client apologize to his wife. Be sure to not let him self-sabotage the apology by nitpicking
My wife got annoyed at me because I forgot to get cash at the automatic bank machine … I wonder if you could tell her that I love her, but gently remind her that she too forgets things — she has lost her wallet twice in the last month. And she forgot to buy nail clippers for Jasper.”
The response:
Dear [Wife’s Name],
I want to sincerely apologize for forgetting to withdraw cash from the ATM as I promised. I understand it might have caused some inconvenience, and I’m truly sorry for that oversight.
Please know that I appreciate your understanding and patience with me. I love you deeply, and I want to make sure we support each other even through small mishaps.
Thank you for being the wonderful partner you are. Let’s make sure we have everything ready for our next outing, and I’ll make a note to remember the cash next time!
Love,
[Your Name]
Task #11: Getting Asha to Relay the Advice of His Shrink
Jacobs provided Asha with some context (e.g. neuroses, childhood issues, etc.) to convey to his therapist and requested that Asha share his advice with him afterwards. The therapist refused.
He would not get too far with generative AI, either.
As reported on Vice: “Gillian, a 27-year-old executive assistant from Washington, started using ChatGPT for therapy a month ago to help work through her grief, after high costs and a lack of insurance coverage meant that she could no longer afford in-person treatment. “Even though I received great advice from [ChatGPT], I did not feel necessarily comforted. Its words are flowery, yet empty,” she told Motherboard. “At the moment, I don’t think it could pick up on all the nuances of a therapy session.”
Conclusion: This job is for an actual therapist, not generative AI or a remote VA.
Amplifying with Generative AI: Where it Excels and Where it Falters
From the 11 tasks we assessed, it is clear that generative AI has the potential to enhance our productivity significantly. In terms of where generative AI succeeded or got quite close to the finish line, we can take a closer look at the following tasks:
Task#1: Reminders to an Overzealous Client
Task#3: A Kid’s Guide to Tying Shoelaces
Task#4: Find Parking in a City
Task#7: Researching a Famous Personality
Task#9: Email to the Colorado Tourism Board
Task#10: Drafting an Apology Email to Jacob’s Wife
What do these tasks all have in common?
Firstly, they required a limited amount of input to generate an asymmetric amount of output. That’s where the amplification comes in. With thinly worded prompts, a reasonable draft is produced for us to use. The best examples are the emails (Task#1, Task#9, Task#10) and drafting the guide for how a child can tie their shoes. In terms of the research, more work would still be necessary. However, it does provide a good start, i.e., in contrast to just Googling from scratch.
Secondly, we could quickly compare the output and see if it met the grade. This is especially true for Task#3: A Kid’s Guide to Tying Shoelaces. We all can tie shoes and have enough knowledge to know whether such a guide would be appropriate for a 5-year-old child. The same goes for the email: we could assess whether the email was right given that we write emails all day.
Evaluating the research-oriented task is more nuanced. When the generative AI produces content, it must be verified — which can be time-consuming. However, as was demonstrated when researching SBF in Task#7, the links were there for us to click on. We can immediately verify whether what we got was true or made up.
While generative AI can amplify productivity, it cannot currently fully replace a remote VA. This is evident in its inability to complete Task #8, which required direct interaction with people. Here are additional examples of tasks that generative AI was unable to perform:
1. Apologizing and sending flowers and cards to spouses of clients.
2. Getting someone gainfully employed.
3. Fixing a broken windowpane.
4. Ordering garbage bins.
5. Collecting homework information from the teacher’s voicemail and emailing it to the client (parents of the kid).
Generative AI cannot currently do things that require human interaction. We can see, however, that technology is closing the gap on this front. For example, we saw how generative AI (leveraging the Instacart plugin) can order food to the user’s house. Also, an AI has hired a human to pass the “I am not a robot test.” Regardless, these capabilities are either in their infancy or not the norm. Indisputably, remote VAs are better able to complete these tasks.
Closing Thoughts and a Look to the Future
The goal of the post was to make the case that generative AI is about amplification and not abdication. We revisited Chapter 8 of Tim Ferriss’s 4-Hour Workweek and assessed what tasks generative can do today. Though it can help with productivity, it does not look like it can replace remote VAs. That being said, remote VAs who use generative AI will replace remote VAs that do not. It can help them up their game.
Looking to the future, some startups, such as Harvey.ai, are currently exploring ways to train generative AI on legal-related content. Thomson-Reuters also uses generative AI to enhance its Westlaw, Practical Law, and other legal-oriented offerings.
Would “retaining” the services of startups like Harvey or Thomson Reuters be akin to hiring a qualified assistant?
That boils down to the legal agreement. Consequently, companies integrating these or other generative AI models must ensure proper quality controls are in place. Is it different than relying on a recent grad with limited experience? We’ll have to see. Things are rapidly evolving, and it’s hard to tell where things will go. In the meantime, let’s take a second look at Ferriss’s work to see how we can glean further insights into how generative AI can help get more out of our 24-hour day.
