Home AI News My two favorite ChatGPT Plus features and the remarkable things I can do with them

My two favorite ChatGPT Plus features and the remarkable things I can do with them

My two favorite ChatGPT Plus features and the remarkable things I can do with them

This text began life as a information to the 2 ChatGPT Plus plugins I depend on probably the most. However as of March 19, 2024, plugins are not accessible for ChatGPT and have been changed by particular person GPTs, which OpenAI calls the ChatGPT market apps. Steven Vaughan-Nichols wrote an in depth rationalization of the transition, so if you wish to know extra concerning the new strategy, I encourage you to learn his piece:

For day-to-day work, I’ve discovered I would like precisely two primary options. 

My two favorites

By far probably the most helpful and dependable function I’ve discovered has been ChatGPT Plus’s knowledge evaluation functionality. This function reads knowledge information, completes evaluation, generates charts, and may even undertakes picture evaluation. (Though I have never used it for that.)

You used to need to activate knowledge evaluation by enabling the Superior Information Evaluation possibility within the ChatGPT Plus Settings dialog. Do not be confused, nevertheless: Superior Information Evaluation was referred to as Code Interpreter. Now, it is simply referred to as knowledge evaluation and you do not have to show it on.

The second high-value plugin I used was WebPilot. This allowed ChatGPT to entry the net. Beforehand, I had been utilizing MixerBox WebSearchG to do the identical factor however discovered it failed as a rule. To date, WebPilot hasn’t failed me.

There have been limitations, nevertheless. WebPilot could not do a lot with complete Amazon searches, for instance. That is as a result of, whereas it may go to and extract content material from webpages, it could not work together with dynamic components on the web page, akin to dropdowns, filters, or search bars.

Now, nevertheless, I take advantage of the embedded net extraction functionality in ChatGPT Plus, which relies on Bing search outcomes. Like WebPilot, the native ChatGPT Plus browser is restricted. For instance, it might’t learn ZDNET pages. That is presumably as a result of web page entry restrictions we placed on the positioning, however it nonetheless can really feel limiting.

Armed with these options, I can do complete knowledge evaluation and guarantee ChatGPT will get present knowledge from the net. After they labored as plugins, I could not use each collectively. Now, I can.

Work inside the limits

There are limits to what these options can do. The browser functionality is nice at pulling in a single web page of data. Typically, as in my Yelp experiment, it might scan a number of pages. This functionality improved when shifting to the inner browser functionality from WebPilot.

This functionality is highly effective as a result of doing a number of net scans after which analyzing these outcomes is the place productiveness good points could be made. Let’s take a latest instance. I am constructing a PC utilizing the AMD Ryzen 5 5600X processor. That processor was chosen particularly as a result of it is really helpful for a chunk of software program I wish to run. I need a motherboard that may host that processor. I attempted to get ChatGPT Plus with the browser to scan Amazon’s choices and critiques comprehensively.

My immediate: 

Use the net. Discover me ten PC motherboards in every value vary that assist the Ryzen 5 5600X processor utilizing the AM4 socket. Embody star critiques and a hyperlink to every product. Present as a desk.

I obtained a listing (it ignored the request for a desk). It gave me star scores. After I used WebPilot, the URLs supplied weren’t to product pages on Amazon. As an alternative, the hyperlinks had been to Amazon’s primary web page. This time, nevertheless, with the built-in browser functionality, it did fairly nicely, offering me with Amazon hyperlinks to particular person product pages.

This strategy is affected by the character of the net, after all. Each web page is coded in another way. Some firms exit of their option to obfuscate web page scraping. Others are troublesome to seize due to how they’re engineered, with none energetic blocking by net builders.

Also, the AI would not like happening a grasp record and presenting outcomes directly. It likes to supply a solution, look ahead to a pat on the pinnacle, after which go on to the following reply. On this manner, ChatGPT is sort of a pet.

The important thing to success is to work inside the limits. For instance, I can choose the motherboards that look fascinating, feed every consequence again to ChatGPT Plus, and ask for sentiment evaluation of consumer critiques — as soon as I find the precise assessment hyperlink for that product.

For instance, I needed to know if one of many boards I used to be serious about had noise points. I had to supply a set of prompts earlier than I obtained some workable solutions:

Do customers have complaints about noise with this product: {insert title of motherboard right here}

This returned an outline of the product, however no sentiment data. So I requested for extra in-depth analysis:

Search the net for that motherboard and noise points

Quite than getting helpful suggestions, the AI really helpful I do my very own legwork:

For extra detailed insights on consumer experiences concerning noise or every other operational features of the motherboard, I like to recommend checking consumer critiques and boards the place people share their private setups and experiences.

I have been discovering that the built-in browser function appears to love returning surface-level data way over in-depth evaluation. Sadly, that considerably nerfs the usefulness of the AI. It looks like the addition of guardrails has rmade the AIs significantly much less helpful on this regard. However then there’s knowledge evaluation.

ChatGPT Plus knowledge evaluation

This functionality is vastly highly effective. After I first seemed, this functionality was restricted to the pre-November 2021 dataset. However now, it would pull in data from the net, with the attention of what I mentioned above. ChatGPT’s net entry is pretty nerfed. Nevertheless, I discovered that if I add my very own knowledge information, the information evaluation functionality shines.

I confirmed a whole lot of charting and desk evaluation utilizing knowledge evaluation on this article. However the place the place importing your knowledge proved extremely highly effective was after I did some sentiment evaluation on a 22,797-record database when uninstalling my code. I hadn’t had the time to put in writing code to do this cross-indexing evaluation, however ChatGPT did it for me in minutes.

I lately used that strategy with one other dataset. There is a service for journalists the place you may put up a query, after which public relations of us and people serious about selling themselves and their firms can reply.

This service is a superb useful resource for getting insights from the parents who’re actively concerned in no matter you are researching. The draw back is the signal-to-noise ratio is large. Many individuals pitch solutions, typically unrelated to what I am researching, and infrequently the individuals pitching do not meet my standards for inclusion in my analysis.

Sifting by means of the pitches for a given mission can take days. I attempted to obtain and feed the total set of responses to knowledge evaluation, however the software did not like that. However after I did some pre-prep work, turning every pitch right into a document in an Excel database (I used a tiny little bit of programming for that job), then knowledge evaluation may perceive the data.

I did not need ChatGPT to do the analysis or evaluation, however it was capable of separate these submissions that included an government’s title and title. I requested just a few key questions that may assist flag an entry from somebody who really learn the instructions. After I requested ChatGPT to parse these entries for these checks, it was capable of current a set of qualifying candidate pitches.

I am not offering these prompts as a result of they had been distinctive to that mission. However the level is that I did some fundamental work to prep knowledge evaluation, after which it did a bunch of research. Normally, it takes me three to 4 days to undergo the tons of of pitches and sift out the workable knowledge. This time, it took me about half a day.

That is an enormous productiveness saving.

The important thing to success with ChatGPT prompts

Definitely, there are limits to ChatGPT, not the least of which is that the AI would not all the time do what you need and it makes stuff up. However ChatGPT is usually a large time saver if you happen to delegate fastidiously.

Discover that phrase: delegate. Used as a verb, Webster’s defines “delegate” as “to entrust to a different”. In case you’re a supervisor, the power to delegate is a vital a part of the job. However delegating is not simply barking assignments at subordinates. Delegating is entrusting, sure. But it surely’s additionally guiding and verifying the work product.

That is how you could be taught to work with ChatGPT. Do not consider it as a pc program. Consider it as an individual to whom you are assigning work. Is that task too complicated? Is that task too obscure? Is there a manner for the particular person to succeed with that task or are you setting them as much as fail?

We’re all accustomed to these office eventualities, so they need to be one thing you may apply to your AI prompting work.

If what you give the AI confuses it, work out the way to make clear. If what you give the AI is an excessive amount of data for it to work with, work out the way to break the mission into phases.

Also, do not restrict your self to 1 software. In case you’re a supervisor, you may need one worker who’s nice at individuals interplay. You may need one other who’s nice at fixing issues, and yet one more with nice math abilities. You would possibly initially assign one mission part to 1 employee after which transfer it to a different as the abilities required change.

That is what I did with my public relations pitch mission. Parsing the replies into comma-separated worth strains was simple for my programmer’s textual content editor. I simply did just a few search and change operations. Then I opened the doc in Excel to see whether or not it was what I assumed I might created. Solely as soon as the information was confirmed did I feed it into ChatGPT and Superior Information Evaluation.

So what’s the important thing to ChatGPT success, particularly with these two options? Study to interrupt the mission down in order that the options can perceive it, after which use all of the instruments at your disposal to get your outcomes.

In case you do this, you may doubtless discover ChatGPT gives a measurable productiveness increase for a few of your initiatives.

Keep in mind that not each mission is one you may give to the AI. Take your wins the place applicable and use different instruments to finish duties unsuited to synthetic intelligence.

Good luck. Dwell lengthy and prosper. Could the power be with you.

You’ll be able to observe my day-to-day mission updates on social media. Make sure you subscribe to my weekly replace e-newsletter on Substack, and observe me on Twitter at @DavidGewirtz, on Fb at Fb.com/DavidGewirtz, on Instagram at Instagram.com/DavidGewirtz, and on YouTube at YouTube.com/DavidGewirtzTV.