WEBVTT

00:00:00.000 --> 00:00:03.500
You've probably heard someone say AI is gonna

00:00:03.500 --> 00:00:07.099
take your job and someone else say AI can't even

00:00:07.099 --> 00:00:11.619
count fingers in a photo both of those reactions

00:00:11.619 --> 00:00:15.519
come from Well the same problem. We're using

00:00:15.519 --> 00:00:19.879
the word AI like it's one Single magic machine

00:00:19.879 --> 00:00:23.039
and guess what? It's not today. I'm gonna give

00:00:23.039 --> 00:00:30.289
you a simple mental model for what what AI is,

00:00:30.969 --> 00:00:35.049
what chat GPT and friends are, and how to get

00:00:35.049 --> 00:00:39.810
the benefits without falling for all the hype.

00:00:40.770 --> 00:00:43.369
Welcome back to the Tyler Woodward Project. I'm

00:00:43.369 --> 00:00:46.649
Tyler, a broadcast engineer by trade, a Linux

00:00:46.649 --> 00:00:49.770
nerd by choice, and I enjoy demystifying tech

00:00:49.770 --> 00:00:54.090
that's supposedly quote unquote too complicated

00:00:54.090 --> 00:00:59.179
for people. Today we're answering a deceptively

00:00:59.179 --> 00:01:03.280
basic question. What is AI and what are chat

00:01:03.280 --> 00:01:08.140
GPT, Claude and Gemini? Here's the plan. First,

00:01:08.200 --> 00:01:12.900
we'll define AI in a way that matches how it

00:01:12.900 --> 00:01:16.680
shows up in your life. Then I'll pull the covers

00:01:16.680 --> 00:01:21.420
off of these chat tools. What they're good at

00:01:21.420 --> 00:01:25.290
and where they get weird. Finally, I'll give

00:01:25.290 --> 00:01:28.670
you a practical checklist for using them safely

00:01:28.670 --> 00:01:32.629
and effectively. One quick note, I'm staying

00:01:32.629 --> 00:01:35.950
platform neutral. The brand names change fast,

00:01:36.069 --> 00:01:40.329
but the underlying ideas are pretty stable. So

00:01:40.329 --> 00:01:52.150
let's get into it. When people say AI, they might

00:01:52.150 --> 00:01:55.719
mean three different things. If you don't separate

00:01:55.719 --> 00:01:59.219
them, everything sounds contradictory to each

00:01:59.219 --> 00:02:03.760
other. First, AI is kind of being used as this

00:02:03.760 --> 00:02:07.140
umbrella term. This includes lots of techniques

00:02:07.140 --> 00:02:10.939
that help computers do things we associate with

00:02:10.939 --> 00:02:15.280
intelligence, recognizing speech, finding patterns,

00:02:15.659 --> 00:02:18.960
making predictions, generating text and so on

00:02:18.960 --> 00:02:23.780
and so forth. Second is machine learning. That

00:02:23.780 --> 00:02:28.139
is when you train a system on examples instead

00:02:28.139 --> 00:02:32.199
of hand coding every single rule. Spam filters,

00:02:32.500 --> 00:02:35.840
photo tagging, and recommendation systems are

00:02:35.840 --> 00:02:40.780
all common examples of this. Number three. And

00:02:40.780 --> 00:02:45.099
this is what's dominating headlines right now.

00:02:45.900 --> 00:02:49.520
Generative AI. That's when a system generates

00:02:49.520 --> 00:02:54.759
new content like Text images audio or code based

00:02:54.759 --> 00:02:59.360
on patterns it learned during its training now

00:02:59.360 --> 00:03:04.680
tools like chat GPT Claude Gemini Co -pilot,

00:03:04.979 --> 00:03:08.259
they're all chat style assistance built on something

00:03:08.259 --> 00:03:14.479
called and an LLM it stands for a large language

00:03:14.479 --> 00:03:18.979
model an LLM is trained on a huge amount of text

00:03:18.979 --> 00:03:23.930
so it can predict what comes next in a sequence

00:03:23.930 --> 00:03:28.629
a useful way to think about it is it's like autocomplete

00:03:28.629 --> 00:03:32.990
but scaled up massively and then tuned to follow

00:03:32.990 --> 00:03:36.050
instructions and and hold a conversation basically

00:03:36.050 --> 00:03:39.990
and it's important to say this out loud these

00:03:39.990 --> 00:03:46.090
tools don't know things the way a person knows

00:03:46.090 --> 00:03:49.229
things they don't have senses they don't have

00:03:49.389 --> 00:03:52.530
lived experience, they don't automatically check

00:03:52.530 --> 00:03:58.650
reality before they answer. What they are, are

00:03:58.650 --> 00:04:03.669
great at producing plausible language. Often

00:04:03.669 --> 00:04:06.710
correct, sometimes wrong, occasionally wrong

00:04:06.710 --> 00:04:09.830
in a way that still sounds overly confident.

00:04:11.069 --> 00:04:15.819
You'll hear a term, hallucination. In this context,

00:04:16.079 --> 00:04:20.100
it means the model generated an answer that sounds

00:04:20.100 --> 00:04:24.879
legitimate, but it's not grounded in real facts.

00:04:25.199 --> 00:04:29.279
It's not lying the way we would think of lying

00:04:29.279 --> 00:04:34.040
in a human sense. It's generating the most likely

00:04:34.040 --> 00:04:39.360
next words. Here's why this matters. If you treat

00:04:39.360 --> 00:04:44.360
a chat bot like a search engine, You're going

00:04:44.360 --> 00:04:47.459
to get burned. Search engines try to point you

00:04:47.459 --> 00:04:51.199
to sources. Chatbots try to produce a response.

00:04:51.459 --> 00:04:54.759
The difference changes how you should trust the

00:04:54.759 --> 00:05:01.500
output. One level deeper, LLMs operate on tokens,

00:05:02.300 --> 00:05:06.139
which are chunks of text. Think pieces of words.

00:05:06.319 --> 00:05:09.439
When you ask a question, the model predicts the

00:05:09.439 --> 00:05:14.649
next token, then the next. then the next, until

00:05:14.649 --> 00:05:18.810
it builds an answer. If you're thinking, so it's

00:05:18.810 --> 00:05:23.149
guessing, yeah, yeah, sort of, but it's guessing

00:05:23.149 --> 00:05:26.990
with statistical models that, you know, it learned

00:05:26.990 --> 00:05:31.870
a ton about how language usually works. This

00:05:31.870 --> 00:05:36.290
reminds me of audio noise reduction and broadcasting

00:05:36.290 --> 00:05:40.639
when it's dialed in, it's magic, but If you push

00:05:40.639 --> 00:05:44.000
too hard, you get watery artifacts that still

00:05:44.000 --> 00:05:48.639
sound smooth. Well, AI techs can be similar.

00:05:49.259 --> 00:05:52.980
Polished output can hide problems. So you need

00:05:52.980 --> 00:05:59.399
a quick reality check. So where do chat GPT,

00:05:59.600 --> 00:06:03.420
Claude and Gemini and all these other chat bots

00:06:03.420 --> 00:06:07.480
fit in? Broadly, they're different products built

00:06:07.480 --> 00:06:11.300
on different model families. with different tuning

00:06:11.300 --> 00:06:15.000
and safety rules and sometimes different add

00:06:15.000 --> 00:06:18.480
-ons like reading files, analyzing images, or

00:06:18.480 --> 00:06:23.519
using web tools. If you've ever tried two assistants

00:06:23.519 --> 00:06:26.279
with the same prompt and getting different answers,

00:06:26.959 --> 00:06:29.540
that's normal. Different training, different

00:06:29.540 --> 00:06:34.180
tuning, different guardrails. The big beginner

00:06:34.180 --> 00:06:37.939
takeaway, don't ask which one is the smartest.

00:06:38.240 --> 00:06:43.860
ask which one fits my task and my risk level.

00:06:45.339 --> 00:06:48.439
Let's connect this to real life with a scenario

00:06:48.439 --> 00:06:51.300
where these tools shine and fail at the same

00:06:51.300 --> 00:06:54.939
time. You've got a pile of information and you

00:06:54.939 --> 00:06:59.879
want clarity fast. Maybe you're writing a cover

00:06:59.879 --> 00:07:03.939
letter. Maybe you're trying to understand a medical

00:07:03.939 --> 00:07:07.660
bill. Maybe you're comparing laptops or maybe

00:07:07.660 --> 00:07:11.839
you're, I don't know, staring at a long email

00:07:11.839 --> 00:07:16.540
thread and you just need actionable items. Chab

00:07:16.540 --> 00:07:20.980
bots are great at turning a messy prompt into

00:07:20.980 --> 00:07:25.240
structure. If you paste a job description in

00:07:25.240 --> 00:07:28.180
your resume and you say, draft a cover letter

00:07:28.180 --> 00:07:32.959
with a confident, friendly tone, well, you'll

00:07:32.959 --> 00:07:36.199
usually get something usable in a few seconds.

00:07:36.560 --> 00:07:39.899
And that matters because it lowers the activation

00:07:39.899 --> 00:07:43.959
energy, if you will. You're not stuck staring

00:07:43.959 --> 00:07:50.000
at a blank page. You can react to a draft instead

00:07:50.000 --> 00:07:53.899
of inventing one from nothing. But here's the

00:07:53.899 --> 00:07:57.279
failure mode. The chatbot will happily fill in

00:07:57.279 --> 00:08:02.430
gaps with confident nonsense. If you say compare

00:08:02.430 --> 00:08:05.389
these two laptops but you don't provide exact

00:08:05.389 --> 00:08:09.170
model numbers and specifications the assistant

00:08:09.170 --> 00:08:13.870
may quietly invent some details for you or it

00:08:13.870 --> 00:08:18.410
might assume you mean a popular model and give

00:08:18.410 --> 00:08:23.730
you a comparison that sounds right but doesn't

00:08:23.730 --> 00:08:29.529
match what you're actually looking for so here's

00:08:29.529 --> 00:08:34.879
a safety a safety pattern. Make the AI stay grounded

00:08:34.879 --> 00:08:38.659
in what you provide and make it show its work

00:08:38.659 --> 00:08:41.879
in a way that you can verify. Instead of asking

00:08:41.879 --> 00:08:45.000
which laptop is better, which is kind of vague,

00:08:45.440 --> 00:08:49.000
ask, here are the specs I'm looking at. Create

00:08:49.000 --> 00:08:53.759
a table of differences using only what I pasted.

00:08:54.120 --> 00:08:59.450
Then tell me which one fits video editing. better

00:08:59.450 --> 00:09:03.049
and why. Now you're forcing it to stay inside

00:09:03.049 --> 00:09:06.570
the fence and you're getting a decision based

00:09:06.570 --> 00:09:09.629
on constraints that you actually care about.

00:09:11.190 --> 00:09:15.690
Another practical example, summarize a long document.

00:09:16.370 --> 00:09:19.250
These tools are excellent at pulling out themes,

00:09:19.370 --> 00:09:22.450
turning paragraphs into bullet points and translating

00:09:22.450 --> 00:09:27.779
dense language into plain usable English. And

00:09:27.779 --> 00:09:30.179
I'll make this personal for a second. For me,

00:09:31.799 --> 00:09:36.360
with ADHD, attention is a limited resource, all

00:09:36.360 --> 00:09:40.639
right? If I'm staring at a long article, a really

00:09:40.639 --> 00:09:44.879
overly dense PDF, or even a messy email thread,

00:09:45.639 --> 00:09:49.600
my brain can bounce off. It's just, it's not

00:09:49.600 --> 00:09:53.779
going to tune in. But if I can paste it into

00:09:53.779 --> 00:09:56.299
a chat bot and say, summarize this into a detailed

00:09:56.299 --> 00:10:00.480
list, break it into sections, and pull out actionable

00:10:00.480 --> 00:10:05.720
items and deadlines, that, for me, has been genuinely

00:10:05.720 --> 00:10:09.879
amazing. It's been, I dare say, life -altering.

00:10:10.059 --> 00:10:13.120
It's really helped. It turned something I'd avoid

00:10:13.120 --> 00:10:15.460
normally into something I can actually start

00:10:15.460 --> 00:10:19.539
with. And starting, that's usually the hardest

00:10:19.539 --> 00:10:24.740
part, right? Now, all this goes without, you

00:10:24.740 --> 00:10:27.879
know, I've... I've got to at least give you some

00:10:27.879 --> 00:10:30.580
caution, all right? So here's where we're going

00:10:30.580 --> 00:10:34.820
to get into the caution part. If the document

00:10:34.820 --> 00:10:40.879
was legal, financial, or medical implications,

00:10:41.100 --> 00:10:45.240
treat the summary like helpful notes, not a final

00:10:45.240 --> 00:10:49.600
product. Verify important points in the original

00:10:49.600 --> 00:10:55.360
text. And if you're making a serious decision,

00:10:55.850 --> 00:11:00.389
Use a qualified professional. One more cultural

00:11:00.389 --> 00:11:05.009
thing that's worth saying out loud. We're trained

00:11:05.009 --> 00:11:09.029
to trust confident language. These tools can

00:11:09.029 --> 00:11:13.490
generate confident language on demand. So we

00:11:13.490 --> 00:11:17.470
have to shift trust away from tone and toward

00:11:17.470 --> 00:11:25.820
verification. Let's turn this into, well, a quick

00:11:25.820 --> 00:11:28.860
beginner checklist you can use today. First,

00:11:29.919 --> 00:11:32.679
tell it what you want, who it should be, and

00:11:32.679 --> 00:11:38.940
what good looks like. Act like a tutor. Explain

00:11:38.940 --> 00:11:43.120
this at a beginner level. Give one example, then

00:11:43.120 --> 00:11:47.919
quiz me. Second, add constraints, so it can't

00:11:47.919 --> 00:11:51.139
drift off. Use only the information I provide.

00:11:52.159 --> 00:11:56.179
If you're unsure, ask me clarify questions before

00:11:56.179 --> 00:12:00.299
answering. I do that sometimes, and it will indeed

00:12:00.299 --> 00:12:02.299
actually ask you stuff back to try to figure

00:12:02.299 --> 00:12:04.679
out what you're getting at, where the end goal

00:12:04.679 --> 00:12:09.399
is. Third, use it for drafts and structures,

00:12:09.740 --> 00:12:13.320
not final authority. It's fantastic for outlines,

00:12:13.659 --> 00:12:17.340
rewrites, and brainstorming. It's risky as a

00:12:17.340 --> 00:12:21.440
pure fact source, unless it can point you to

00:12:22.340 --> 00:12:26.159
verifiable references you can check. Fourth,

00:12:26.919 --> 00:12:31.779
learn the red flags. If it gives you very specific

00:12:31.779 --> 00:12:35.320
numbers, quotes, or according to a study, but

00:12:35.320 --> 00:12:38.019
doesn't give you the source of what that study

00:12:38.019 --> 00:12:42.820
actually is, treat that as a cue to verify. If

00:12:42.820 --> 00:12:46.480
it can't provide a grounded answer, you can ask

00:12:46.480 --> 00:12:50.539
it what it would need to know or what sources

00:12:50.830 --> 00:12:56.409
you should consult. And fifth, protect sensitive

00:12:56.409 --> 00:12:59.450
information. Assume anything you paste could

00:12:59.450 --> 00:13:02.009
be stored depending on the service and settings.

00:13:02.649 --> 00:13:07.149
Don't paste passwords, private keys, or confidential

00:13:07.149 --> 00:13:11.230
work documents unless you explicitly know your

00:13:11.230 --> 00:13:15.610
workflow is approved and safe. Don't get yourself

00:13:15.610 --> 00:13:19.909
in trouble. That's the point. used well these

00:13:19.909 --> 00:13:22.730
tools can save time and reduce friction used

00:13:22.730 --> 00:13:27.029
carelessly they can quietly inject errors into

00:13:27.029 --> 00:13:31.470
important decisions so the next time someone

00:13:31.470 --> 00:13:35.490
says AI you can ask do you mean machine learning

00:13:35.490 --> 00:13:40.250
in general or do you mean a generative chatbot

00:13:40.250 --> 00:13:44.549
an LLM that's great at language but not automatically

00:13:44.549 --> 00:13:49.320
great at truth visit tylerwoodward .me, follow

00:13:49.320 --> 00:13:51.960
at tylerwoodward .me on Instagram and threads,

00:13:52.480 --> 00:13:55.440
subscribe and like the show on your favorite

00:13:55.440 --> 00:13:58.240
podcast platform. I'll catch you next week.