back to indexBing Chat (GPT 4) Tutorial: 12 Steps for Beginners to Advanced - Crash Course!
Chapters
0:0 Intro
1:18 Step 3 Link Articles PDFs Academic Papers and Physical Text
2:50 Step 4 Generate Interesting Ideas
5:34 Step 5 Bing Search
7:0 Step 6 Bing Shopping
7:50 Step 7 Writing
9:5 Step 8 Creative Writing
10:50 Step 9 More Advanced
12:18 Step 10 Coding
13:14 Step 11 Aggregation
14:16 Step 12 Tutor
15:3 Bonus Step
00:00:00.000 |
Bing Chat is an extraordinary tool that yes can make mistakes but which can also do amazing things 00:00:06.240 |
and give you a real edge in whatever you want to do. This crash course will take us from complete 00:00:11.840 |
beginners up to advanced users following 12 steps each of which is potentially more useful than the 00:00:18.800 |
last. Step one starts by us going to bing.com and clicking chat. You'll see this screen with a chat 00:00:25.840 |
input down below. If you don't see this and don't yet have access to the chatbot feature my tip is 00:00:31.600 |
to follow the four instructions on the bing.com/new website which are to join the waitlist, set 00:00:37.760 |
Microsoft defaults and install the bing app. You can easily undo these steps if you like once you're 00:00:43.120 |
in and I gained access within about 48 hours. Alternatively to be honest bing chat access will 00:00:48.880 |
be rolled out to everyone eventually. But before we can begin on what bing can do step two involves 00:00:54.480 |
knowing the limits of bing. If you're a beginner you can start with bing chat and then you can 00:00:55.820 |
start with bing chat. As of filming you have a 2000 character limit on inputs, a five message per 00:01:02.600 |
conversation limit and a 50 message per 24 hour period cap. And one further interesting restriction 00:01:09.600 |
to point out bing really doesn't like it if you try to talk to it about itself. It doesn't even 00:01:14.960 |
like it if you ask it for its name. I've done a video on this. But now let's get to the good stuff. 00:01:19.440 |
Step three you can link to articles, pdfs, academic papers and even physical 00:01:25.800 |
text and engage in a dialogue with bing about what's inside. This is truly revolutionary. 00:01:32.500 |
I linked to an article in today's new york times about how leonardo da vinci had thoughts about 00:01:37.780 |
gravity even before newton and I simply asked eli5 explain it like i'm five. Of course bing 00:01:44.260 |
understood that instruction but what's truly incredible is that it was able to read and 00:01:49.340 |
digest an article from today, explain it in really simple language, try reading it yourself. 00:01:55.780 |
Just look at the details it picks out. I could rave about this for ages I've just got so much more to 00:02:00.900 |
show you. But what did I mean by engaging with physical text? You can take a photo of anything, 00:02:05.740 |
for example this math problem. Send it to your desktop and then go to google image search. Drag 00:02:11.500 |
the file onto the screen and this will happen. Click on text to extract the text. Grab the text, 00:02:18.520 |
in this case the question and you can guess what's coming next. Yes that's right go to bing and paste 00:02:25.060 |
the question in. Now you can see that bing has a lot of information about the text. It's a lot of 00:02:25.760 |
information. I know the meme is that GPT powered models aren't good at math but as a math tutor I 00:02:30.480 |
can say that's old news. I've done plenty of videos on this on my channel, check them out. 00:02:35.280 |
But AI is getting a lot better at math fast. It gets this question right with a nice explanation. 00:02:40.880 |
But imagine this for physical text that you want to explore in your environment. 00:02:44.960 |
Three steps from the camera to a conversation with a super advanced AI about it. Incredible. 00:02:50.640 |
But it's time for the next step, generating interesting ideas. This could be for youtube, 00:02:55.740 |
instagram, tiktok, linkedin, whatever. To demonstrate I asked, generate five original 00:03:01.180 |
youtube video ideas for a new consumer tech review channel. Detailing the novelty of each one, 00:03:06.620 |
ensure the topics are trending in 2023 and write a detailed synopsis of each video with a hook, 00:03:12.380 |
a challenge and a reveal. Feel free to steal any of these prompts by the way. 00:03:16.540 |
And you can read the answers for yourself but I think they're pretty interesting. Any one of 00:03:21.020 |
these would make a viable video idea, in this case for youtube. Let's read the first one, make a video 00:03:25.720 |
about the mac mini 2023, the ultimate desktop computer question mark. And the video idea is to 00:03:30.820 |
test the performance, the design, the features, talk about the m2 chip and how to optimize the 00:03:36.060 |
mac mini experience. It gives specific examples of the tech that you could compare and suggestions 00:03:41.820 |
of how to make your video different, like testing them out in different scenarios such as in the gym 00:03:46.980 |
or at home. What about twitter? It can do that amazingly too. I follow Tim Urban, always coming 00:03:52.500 |
up with new interesting ideas and I said write five real interesting ideas for youtube. Write 00:03:55.700 |
down your original tweets in the style of Tim Urban. And these tweets were incredible. You can 00:04:00.000 |
read all of them but take the fourth one. There are more stars in the observable universe than 00:04:04.520 |
grains of sand on all the beaches on earth. But there are also more atoms in a single grain of 00:04:09.160 |
sand than stars in the observable universe. So which is bigger, the universe or a grain of sand? 00:04:14.680 |
I can imagine him asking that. But let's take this a step further by bringing in image generation. 00:04:19.440 |
I asked Bing, generate five incredibly descriptive visual mid-journey prompts on topic four. 00:04:25.680 |
It understood exactly what I meant and came up with these examples. I put them into mid-journey 00:04:30.300 |
and here's what came out. A grain of sand magnified to reveal a complex microcosm. I think the third 00:04:36.300 |
result is best here. Or what about the universe as a tiny speck inside a giant eye that belongs 00:04:41.500 |
to an unknown cosmic entity? Look at the second example. These would definitely be eye-catching 00:04:46.760 |
in a social media post. Just quickly, here are some of the others. A surreal collage of different 00:04:55.660 |
images of the earth and the sun. I picked out just a couple of these examples to upscale into a full 00:05:03.800 |
image with even more detail. And here was one of the results. That is an AI image you can attach to 00:05:10.760 |
a tweet written by an AI. The world is getting really weird. Before we move on to the next step, 00:05:15.760 |
let's not forget about the 80/20 principle. Yes, we might let Bing AI do 80% of the work, 00:05:21.760 |
but you still have to fact check and make these posts, these ideas, 00:05:25.640 |
even these image generations your own. I wouldn't recommend letting Bing do all the work. And I would 00:05:30.620 |
definitely check its outputs. It does make mistakes. Next, what about using Bing search to actually, 00:05:36.740 |
you know, search? Imagine you're looking to move home and look at the kind of search you can do 00:05:42.020 |
with Bing AI. Compare property price trends, commute to King's Cross times, crime rates, 00:05:48.020 |
median ages, and green spaces in the two London boroughs of Haringey versus Barnet. And here are 00:05:55.620 |
And of course, I could continue this conversation by asking more about green spaces, median age, 00:06:02.040 |
etc. Compare this result to Google. Now, I still do use Google and will continue to. But honestly, 00:06:09.000 |
these results are just not too useful. I would have to click on multiple links to even have a 00:06:15.420 |
hope of coming up with some of the answers that I've already gotten. Let me give you two more 00:06:19.560 |
examples to balance it out. I asked Bing list five electric car charging points nearest to Big Ben in 00:06:25.600 |
London. And it gave me five charging points. And these do check out. They are near Big Ben, 00:06:31.240 |
and they are suitable. But they are all from Q Park. And Google gave me more varied results. So 00:06:38.440 |
I count that as a win for Google. Another win for Google came when I asked, where can I buy flowers 00:06:44.440 |
nearest to the second option? It understood what I wanted, but gave me some online options. Even 00:06:49.780 |
when I was more specific and said I need a physical store florist within 15 minutes walk, it gave me some 00:06:55.580 |
pretty poor results. Google did much better. How about doing a search for shopping? Well, 00:07:01.100 |
I asked for the best battery cases for the iPhone 14. And it gave me some good options. And then I 00:07:06.740 |
was more specific. I want the biggest battery capacity and it must be under 30 pounds. These 00:07:12.620 |
were some decent results, but I would have hoped for direct links for purchasing these cases. 00:07:17.660 |
Remember, every click counts in the war between Google and Bing. 00:07:21.320 |
Even when I said compare these two cases for me, which is what Bing suggests, 00:07:25.560 |
I said, it did give me this nice table, but again, no direct links. And one extra slight warning. It 00:07:32.700 |
suggested a case that isn't actually available in the UK, despite me making it fairly clear that I 00:07:39.000 |
was from the UK by asking for the price in pounds. And of course, my IP being in the UK. So in terms 00:07:45.360 |
of search, the wind still goes to Google, but Bing chat does have its use cases. 00:07:49.740 |
The next step to being an advanced user is to use Bing AI to improve your writing. A 00:07:55.540 |
student of mine wrote this as their introductory paragraph for their personal statement, 00:07:59.800 |
a bit like a cover letter for a university. It's not bad, no spelling mistakes, but the writing is 00:08:05.140 |
kind of bland. And look what Bing was able to suggest. These are nuanced comments, a step 00:08:11.440 |
change from ChatGPT. If you want to learn more, by the way, about how Bing AI is smarter than 00:08:16.960 |
ChatGPT, I've done two videos on this topic on my channel. Let's look at a couple of the suggestions. 00:08:22.240 |
It says the introduction is too long and verbose. 00:08:25.520 |
It could be shortened by removing unnecessary details and using more concise language. 00:08:30.380 |
For example, instead of saying, I have always been an observant and enthusiastic individual, 00:08:34.700 |
always looking for answers to how and why things work in a particular way, 00:08:38.480 |
which is what my student wrote. You could say, I have always been curious and eager to learn how 00:08:43.400 |
things work. Much shorter, a great suggestion. Bing goes further and actually rewrites the 00:08:49.520 |
introduction. I have read this, you can too, and it is a significant improvement. 00:08:53.420 |
I mean, you could improve it still further, but that 00:08:55.500 |
would take real writing skill to do. But notice it didn't just spit out the output. It gave reasons 00:09:01.080 |
for each of its suggestions. And that's vital if you want to learn and improve. 00:09:05.160 |
Next step is Bing's ability to tell jokes and do creative writing. This is kind of a hit and 00:09:11.400 |
miss affair. It's quite hard to get Bing to do what you want, but when it does, it works really 00:09:16.380 |
well. Notice what I tried. I said, write five actually funny original jokes, and it found 00:09:22.560 |
jokes from the web. Now, I think some of these are quite funny. For example, 00:09:25.480 |
what do you call cheese that isn't yours? Nacho cheese. But these aren't original. It stole these. 00:09:31.660 |
So I said, write five jokes about Twitter. And the first one's decent. Again, these aren't Bing jokes, 00:09:37.780 |
though. They're stolen. First joke is, a man tells his doctor, Doc, help me. I'm addicted to Twitter. 00:09:42.100 |
The doctor replies, sorry, I don't follow you. Not bad. Three of the jokes, by the way, aren't even 00:09:48.160 |
about Twitter. Not sure what's going on there. But then when I asked it to do a comedy skit, I 00:09:53.200 |
initially faced some problems. I was told that I was going to be a doctor. I was told that I was going 00:09:55.460 |
to be a doctor. I was told that I was going to be a doctor. I was told that I was going to be a doctor. 00:09:55.520 |
The first two times I asked, it simply refused. It said that topic is too sensitive and 00:10:00.880 |
controversial for me to joke about. It was going to be about TikTok, Donald Trump, Elon Musk. 00:10:05.000 |
I tried again, removing the brand name of TikTok, and just said social media, and again it refused. 00:10:10.620 |
However, I tried one more time and made it fully generic, and it picked up on what I actually 00:10:16.500 |
wanted and wrote something great. I asked, somewhat generically, write a funny and creative 00:10:20.620 |
story about a billionaire buying a big social media company and using it to promote his posts. 00:10:25.440 |
This time it complied. And you can read it for yourself, but I think it's quite entertaining. 00:10:29.160 |
Elon Musk was bored. He had already conquered space, electric cars, tunnels, and brain chips. 00:10:33.580 |
He wanted a new challenge. He wanted to own Twitter. It goes on and on, and then finally 00:10:39.780 |
ends with, Twitter became Elon Musk's playground, and nightmare, and empire. The end. So it can do 00:10:46.920 |
creative writing if you prod it in the right way. In the next step, more advanced now, you can use 00:10:55.420 |
Your analysis assesses the strengths, weaknesses, opportunities, and threats of a given business. 00:11:01.060 |
And Bing was able to do this for a startup I chose almost at random, Hugging Face. 00:11:05.600 |
It's answers are detailed and interesting. Imagine this applied to your business. And 00:11:10.500 |
of course you can continue the conversation or follow the links to find out more. 00:11:14.740 |
Next, it could help on a more granular level. You can use it to respond to reviews. 00:11:19.780 |
This is a real review, two stars, posted about the Peninsula Hotel in Paris. I'm going to 00:11:25.400 |
let you compare the response that Bing provides with the actual response that the Peninsula 00:11:30.380 |
Hotel Paris wrote on TripAdvisor. Essentially this person was complaining about lots of 00:11:34.960 |
things, mentioning some positives, but complaining about quite a few things as well. And my prompt 00:11:40.960 |
was, write a polite, professional, and detailed response to this review. I think this response 00:11:47.260 |
is incredible. It goes into detail, addressing each of the points. And this was a mind blown 00:11:52.260 |
moment. It actually uses a real email. I checked this. 00:11:55.380 |
I'm going to read this and compare it to the official response. 00:12:02.440 |
I mean the response is professional and it is polite, but it's far less detailed. There 00:12:07.220 |
are some grammar issues and it doesn't address all of the points. If you could have a professional 00:12:12.160 |
response ready instantaneously for any of your customer reviews, wouldn't that help 00:12:17.080 |
your business? Next, you can use Bing AI to get into, or to 00:12:21.140 |
get better at, coding. I'm a beginner at coding, but it was able to help me get better at coding. 00:12:25.360 |
I think it's a great way to get better fast. Take this simple request. I've experimented 00:12:29.660 |
with probably a hundred pieces of code, but I just wanted to give you a simple example. 00:12:34.120 |
I asked, write Python code that will add 5% to a user's number and then divide by three. 00:12:40.340 |
This is a simple math request that you can imagine being used for a restaurant, for example. 00:12:44.460 |
If you're new to code, you can always try this out in an application like VS Code, or 00:12:48.940 |
just run the code online as a fun experiment. I pasted it into online-ide.com, for example, 00:12:55.340 |
which has quite a few coding languages to choose from, and then you press run. It will 00:13:00.960 |
prompt you for a number. I entered a number and it gave me the correct result. Only your 00:13:06.380 |
imagination can limit you in terms of what you can try out with coding, even for complete 00:13:11.360 |
beginners, now that you have Bing and ChatGPT. 00:13:14.840 |
Next step, you could use Bing Chat as a quick and easy review aggregator. Only if things 00:13:19.980 |
you don't care too much about, for example, like TV shows, movies, restaurants maybe, anything particularly 00:13:25.320 |
important, you want to check the reviews yourself. But here's what it came up with 00:13:28.940 |
when I asked it for IMDB Rotten Tomatoes Metacritic Empire and Guardian reviews of The Last of 00:13:36.220 |
Some of these numbers are a little off because I think it got confused with the game. And 00:13:40.320 |
even when I clarified, it didn't give the most comprehensive result imaginable. But 00:13:45.260 |
for quick and easy review aggregation, I think it might get really handy. Same for testing 00:13:50.240 |
out any worries you might have about a restaurant or hotel. For example, I asked, are there 00:13:55.300 |
any negatives for staying at the Peninsula Hotel Paris? I feel like I'm picking on that 00:13:58.800 |
hotel. It is one of the best hotels in the world apparently. I'm definitely not trying 00:14:02.180 |
to pick on it. But Bing was able to find some negatives that reviewers have pointed out. 00:14:08.100 |
This saved the time of me scrolling through the reviews to find out what might be a slight 00:14:12.620 |
issue. Skip straight to the positives and the negatives. 00:14:16.580 |
Next step, simply using it as a tutor. And I say this as a tutor. 00:14:20.680 |
But I certainly couldn't find you five peer reviewed studies on the link between career 00:14:25.280 |
team consumption and cognitive performance, nor could I summarize the findings instantaneously. 00:14:31.300 |
Here are the results. And of course, I could continue this conversation and ask about the 00:14:35.400 |
molecules inside creatine, the sources of it, the cost of it, other reasons for taking 00:14:42.800 |
Or let's say you don't care about creatine and you want to know about large language 00:14:46.820 |
models, the models behind Bing AI and ChatGPT. You can use Bing to teach you. It can't quite 00:14:55.260 |
But it's free and can give you a great start on a topic. It's also never grouchy and doesn't 00:15:04.020 |
Finally a bonus step. You can use it to improve your own health. I said, write me a high protein 00:15:09.580 |
vegetarian meal plan for a week and give me five tips on how to actually stick to it. 00:15:14.860 |
The tips were decent. So I followed up by saying, give me this as a daily meal plan. 00:15:20.440 |
Include protein shakes. It gave me a one day meal plan and even told me how to do it. I 00:15:25.240 |
thought that was a great idea. But it was not. So I decided to do a few simple exercises. 00:15:28.060 |
Think of it more as a source of inspiration rather than a final authority on any matter. 00:15:34.460 |
If you think of it like that, Bing's utility is almost endless. 00:15:38.060 |
If you want to find out more about just how smart the new Bing is, check out my Bing chat 00:15:43.720 |
playlist. I detail Bing's current IQ. It's sometimes surprising conversations. 00:15:50.120 |
If you check out my GPT-4 playlist, you can find out about 8 upgrades that may be coming to 00:15:55.220 |
Bing AI in the near future. If you found these steps at all useful, please do let me know 00:16:00.820 |
in the comments. It really motivates me to make more such videos. Have a wonderful day.