AT&T Presents - Winning in a GenAI Disruptive World

Summary

The speaker reflects on previous technology disruptions like the internet and mobile apps to provide context on the disruption caused by AI models like ChatGPT. He explains how ChatGPT is being rapidly adopted across industries and shares examples of how his company AT&T is using it to boost productivity. He advises graduates to learn these new skills to stay competitive in the workplace or consider entrepreneurial opportunities building on top of language models, emphasizing key principles like iterating quickly, enabling broad use cases, and automating actions.

Takeaways

• 😲Generative AI like GPT is massively disruptive, touching many industries and being adopted very quickly
• 👩‍💻As a new grad, get hands on with generative AI fast to stand out and be productive
• 🤝Enable yourself broadly with generative AI - automate workflows, analyze data, write content etc.
• 💡Build on top of large language models for the most value - the models themselves may commoditize
• 🔀Iterate applications of generative AI - it won't be perfect at first, incorporate human feedback
• 🚀For entrepreneurs, now may be the best time ever - huge opportunities in this disruption
• 💰Get to generative AI applications fast for outsized gains - it's poised to be a 'winner takes most' market
• 🤖Go beyond information to recommended actions and automation for the highest value from generative AI
• 📈Legacy workers may struggle - as a young grad you likely have big advantages
• 😎Don't be narrow in your thinking - small generative AI teams will build huge businesses

Q & A

What are the three major technology disruptions the speaker has witnessed in his career?

The three major technology disruptions the speaker referenced are: 1) The rise of the internet, the World Wide Web, search engines, and email in the 1990s; 2) The emergence of companies like Amazon and the shift to online retail around 2002; and 3) The launch of the iPhone in 2007 which dramatically increased mobile data usage and access.

How does the speaker explain the wide-ranging impact of generative AI models like ChatGPT?

The speaker shows data indicating generative AI has potential impact across many sectors - finance, customer service, education, healthcare, etc. He says it 'almost affects everything' and seems more disruptive than previous innovations because of its broad applicability.

What does the speaker mean when he advises people to 'get there fast' regarding generative AI?

He means that early adopters of new technologies like generative AI often gain a dominant market position, using the examples of companies like Google, Amazon, and Apple. His advice is to start implementing and leveraging generative AI as soon as possible before competitors do.

What does the speaker mean by 'enable yourself broadly' in terms of preparing for disruption from generative AI?

He advises developing skills across multiple roles rather than specializing, because generative AI may allow automation of certain specialized roles. His example is a developer possibly taking on other responsibilities thanks to productivity gains from generative AI.

Where does the speaker suggest focusing for the most value on top of generative AI models?

He suggests building capabilities on top of the core language models, rather than trying to differentiate solely with better language models. Examples could include domain-specific implementations or going beyond information retrieval to take automated actions.

What unique advantages does the speaker highlight for students preparing to enter the workforce?

He notes that as digital natives, students are adaptable to new technologies like generative AI. They can take advantage of more experienced coworkers who may struggle adapting to innovations like chatbots and AI assistants.

When does the speaker suggest entrepreneurial opportunities may be greatest with regard to generative AI?

He proposes that the current period, with rapid advances in generative AI across industries, represents an unprecedented opportunity for enterprising startups to harness these tools to disrupt established players.

What AT&T-specific uses does the speaker describe for generative AI models?

AT&T is using generative AI for internal Q&A, documentation, accelerating workflows, generating code, classifying/translating documents, creating marketing content, etc. The speaker mentions they had 260 initial AI use cases submitted internally.

How did AT&T attempt to manage the flood of generative AI use cases within the company?

After 260 use cases were submitted in just 3 weeks, AT&T had business leaders vet ideas and nominate the most valuable for implementation rather than attempting to pursue all proposals.

What does the speaker's RHF acronym stand for regarding best practices with generative AI?

RHF stands for: 1) Reinforcement learning: improve through a feedback loop 2) Human feedback: incorporate input from users

Outlines

00:00

😄 Introducing chatGPT and discussing technology disruptions

The speaker introduces chatGPT, noting how it seems disruptive and is being rapidly adopted. He reflects on previous technology disruptions he has witnessed in his career in technology, such as the rise of the internet, e-commerce, smartphones, and mobile apps. He emphasizes how generative AI like chatGPT has potential for broad impact across industries.

05:03

😲 Generative AI use cases at AT&T

The speaker discusses the wide range of generative AI use cases at AT&T across business functions like customer service, coding, education, healthcare, marketing, translation, document classification and summarization. He notes the rapid influx of 260 proposed use cases in just 3 weeks, more than they can handle currently.

10:04

🤔 Reflecting on lessons from past disruptions

In light of the rapid changes, the speaker reflects on lessons he has learned from previous technology disruptions during his career. He advises the audience to "get there fast" and adopt new skills early, to think broadly across domains in how they apply skills, to build something novel rather than rely solely on existing models like chatGPT, to move from information to recommended actions to automation for greater impact, and to iterate based on user feedback.

Keywords

💡Disruption

A disruption is a major change that significantly alters how industries, businesses, and consumers operate. The speaker discusses several major technology disruptions he has witnessed, like the rise of the internet, e-commerce, smartphones, and now AI. He argues generative AI like ChatGPT is uniquely disruptive because of its rapid adoption and broad impact across sectors.

💡ChatGPT

ChatGPT is a new generative AI system released in 2022. It can generate human-like text and converse based on prompts. The speaker highlights how quickly ChatGPT gained users, arguing its speed of adoption makes it uniquely disruptive. He also shows ChatGPT's capabilities through AskAT&T, an internal AI assistant.

💡Adoption

Adoption refers to how rapidly new technologies spread through industries and consumer bases. The speaker contrasts ChatGPT's extremely fast user adoption to other disruptive technologies, which took longer like Netflix. This signals the disruptive potential of generative AI.

💡Enable

The speaker advises students to "enable yourself broadly" with AI skills instead of specializing. As AI automates narrow specialty roles, versatility will be valuable. Students knowing AI may even expand their capabilities to do more types of work.

💡Automate

The speaker argues the key to winning with AI is to not just get information from it like ChatGPT answers, but automaterecommended actions. Fully automating workflows with AI will create even more value. This continues the theme of AI's disruptive potential.

💡Iterate

Iterate means trying something, getting feedback, and improving on the next try. Since AI and language models won't be perfect immediately, the speaker advises iterating using customer feedback to continually improve AI tools.

💡Language models

Language models are AI systems trained on vast text data to generate readable, human-sounding text - key to systems like ChatGPT. The speaker warns students not to build businesses solely around large language models as they may become low-value commodities.

💡Entrepreneur

With AI as a disruptive force, the speaker suggests students have a unique opportunity for entrepreneurship. By harnessing language models and carving out valuable niches, small teams could build hugely impactful businesses.

💡Workforce

The speaker notes how some corporate workforce members are less adaptive to new technologies like AI versus students. By gaining AI skills now, students can outpace and replace slower-learning legacy workers.

💡Skills

The speaker frequently advises students to develop AI skills, as they are young and adaptable. He argues AI skills will make students productive and competitive in corporate jobs or better able to start their own companies.

Highlights

ChatGPT seems disruptive and is being adopted very quickly

ChatGPT appears to impact many industries and areas of business

AT&T is using chatbot technology like ChatGPT for many internal use cases

Chatbot assisted teams at AT&T were 20-50% more productive in trials

260 initial AI use cases came in from AT&T business units in just 3 weeks

Being an early adopter of new technologies can provide a competitive advantage

Broadly enable yourself with chatbot skills instead of specializing narrowly

Large language models like ChatGPT may become a commodity, focus on novel applications

Consider going beyond information to recommended actions and automation

New graduates have unique advantages to adopt chatbot technologies

Now may be the best time ever to explore chatbot entrepreneurial opportunities

Join AT&T data science programs to get hands-on chatbot experience

Iterate chatbot applications based on user feedback for continuous improvement

The next major tech company may only need 10 people because of chatbots

English is becoming the new code, you can generate apps by describing needs in plain language

Transcripts

[Music]

so I was trying to think about what am I

going to talk about what do I think

about what's happening there's been two

talks brought up chat GPT I'm going to

give you a third one okay what do we

think is actually happening with that

ever since chat GPT came out earlier

this year it's like oh my gosh how do I

survive how do I win in this world what

am I going to do so I started thinking

well this thing looks

disruptive um well wait a minute I've

seen some disruptions I mean I'm old I'm

from three technology disruptions ago

but I'm not that old 1890s still have it

right but I have seen three internet

technology disruptions the first one I

saw is when I graduated I'm the only

Georgia Tech one here about the 1990s

the internet was blowing up worldwide

web search happened

email it was so strange when I first

started Bell South I was one of the

first people to get email I was like

what do you mean I had it at Georgia

Tech you're just getting it now and then

all of a sudden 2002 retail started I

see my colleague from Amazon here right

Amazon came on the scene the cloud

started happening another disruption and

then I was in Wireless operations I was

trying to plan this the the cellular

network and we thought we're doing fine

until the iPhone came out data blew up

right overnight 2007 we couldn't put in

radios fast enough in the towers like it

was hard to keep up why was that they

made it usable they gave us mobile apps

social happened it was a big thing it

was a big big disruption so today I

happen to be fortunate to be in the

chief data office I can do Ai and data

science generative AI is blowing up I'll

tell you 80% of everything we do is

generative AI chat GPT type stuff and I

was trying to think of disruption I

couldn't think of this whole work from

home thing I could not work from home

because of those disruptive dogs Milo

and Quincy okay but I'm not talking

about that today I'm talking about

generative AI now disruption what do we

mean by that there's the definition

there right something that significantly

Alters the way consumers Industries and

businesses operate and I showed you

these things what are the companies that

did this Google right disrupted search

disrupted advertising

Amazon you don't buy books you know in

the book store anymore you go bid them

online but all this thing online retail

we talked about Apple with smartphones

but notice these things they're kind of

in an area right and if you think about

generative AI who's going to be the

winner there maybe Nvidia maybe an open

AI I think they were valued at 90

billion but what's different is that

question below it feels different it

feels like it's almost affecting

everything right now what else is

different

the adoption is so fast have you seen

this Chad GPT two months 100 million

people right Netflix was 18 months look

at that right the fastest ever now

here's the thing that blew my mind we

look at it today it almost affects

everything this is the percent impact

right what scared me was like the first

one wait a minute that's me D science

right right customer care programming

you're probably using it to write code

right hack code fix code right we use it

all the time education you know sitting

talking about me do my homework right

there right healthare everything right

percent it's like it's everywhere this

is disruptive right now I want to give

you a feel for AT&T okay I told you 80%

of the stuff we're doing is that now let

me show you some of the use cases right

Chad gbt is great you get answers but

does knows nothing about AT&T I can't

ask it how many days off do I get if

I've been here 5 years at 18t isn't no

right I can't ask ask a finance question

can't ask a network question so we have

so many use cases I want to do I want

those answers I want them fast I don't

want to search for it I just want to ask

a question so we're enabling all these

business units to ask questions by

teaching it about

AT&T on top of that we have the outof

the boox thing coding right generate

some code fix this code migrate this

code we did a trial we put one group you

said you can't use ask at we call chat

gpts ask AT&T internal it's our own

private instance we said you can't use

it you can the people that used it 20 to

50% faster depending on who you talk to

okay huge productivity gain but it's not

just data it's not just documents it's

not just code now we have English as the

new code we just write it I don't have

to write SQL I say tell me how many

customers post bait I had last month

boom there it is right so it's amazing

we have all these other use cases

marketing create me some content right

uh translate this from English to

Spanish we have so many customers that

that speak Spanish they call us all of

our documents are in English they have

to translate on the fight why don't I

just do that for them right classifi

documents summarized documents now to

give you some feel for how disruptive

this is we put out something said tell

us your use cases and in three weeks we

had 260 use cases across all the

business units we had to shut it down we

couldn't we couldn't take them all so we

said is okay it's too many we're going

to nominate Business Leaders you vet

them and you bring them to us and then

we'll figure out which are the most

valuable ones now what does that mean to

me what does it mean I'm on the top of

the list there right and you should

think about it too it's like hey I'm

graduating what does that mean to me

what if I go in the corporate world the

other thing you might think about is is

this an opportunity to be an

entrepreneur so I kind of think about

those those two personas as we go

through this here now I did a lot of

thinking is like what have I learned

of these past disruptions and what have

I learned at AT&T that maybe gives some

Clues as to how might you win in this

world and it happens to actually spell

gen the first one is the first principle

is get there fast okay no matter what I

see Sydney had a great example she came

up with this great idea we're the first

in line with the great idea we get

funded okay the same thing happens right

get there fast right so think about if

you're coming out you're going into AT&T

or the industry you're coming out with

some of these skills you're going to be

in the in this workplace some people

have been there 20 30 I had a person in

my group was there 54 years okay some of

them might not want to learn the new

thing right you're coming out you can

take advantage of that right because

you're going to be sitting let's say you

go into coding right and this person

next to you let's say they do it and you

don't they're going to be 50% faster

right you got to have it to compete now

what if you're an Entre you want to do

entrepreneur thing you if you look at

these big disruptive things what it

looks like is if you get there fast it's

almost a winner take all if not most

right look at the market share of those

folks right AT&T Apple dominates they

dominate everywhere in smartphones right

so think about that also right get there

fast um for both cases no matter which

way you go now the next one is enable

yourself broadly what do I mean by that

I'm showing you ask ATT here this is our

own internal Chad gbt I just ask the

write some code that's good for you but

could you actually do something more now

let's say you're in that developer role

that second role right there right if

this thing is so powerful you could be

more productive but you actually might

be able to do some of those other roles

too so it might not look like that it

might look like this right you can

possibly pick up some of those other

things in fact they say the next billion

doll company in the next few years might

only be 10 people why because of this

thing okay so don't think you're narrow

in fact we're doing it now we actually

crowdsource the requirements from all

the comments we put in chat gbt all the

comments we create requirements and then

we create code directly from them just

cut out that whole thing right so super

powerful look to the left look to the

right now of course if you're going to

do something you got to have features

got to be something new what is that

going to be in this gen world is it

going to be these large language models

Chad GPT is that it let's say Chad GPT

is off to the right there that's the

reference okay did you know that when

they put out the open source in 3 weeks

they were 92% of the quality of chat GPT

that might be actually a commodity at

some point so what are you going to

do well let take the lessons from the

internet I think AT&T is 20% % Victor

will correct me 20% of the world's

traffic we carry it's a bit of a

commodity right look at where all the

value was on top all those folks got the

dollars there right it might be the same

thing here the large language models

might be a commodity think about doing

something on top of them like Sydney had

Ops GPT did it on top of it I didn't I

wasn't limited by it right so that's

where the value is now this one is maybe

not so obvious let me try to explain it

automated ACC

now chat GPT is great you get

information okay information from chat

GPT that's cool information about your

own stuff that's even better right what

if you get information in recommended

actions okay now I can actually do

something what if you automated those

actions even more value right let me

just give a quick example summarize this

public data okay chat GPT summarize my

meeting we're doing it today you take

the transcript you automatically

summarize the meeting boom I don't have

to take notes what if I create all the

action items too that's even more

valuable what if I automated the action

items send this email set up the next

meeting do this reboot this thing right

that's where the value is so think of

the full thing go to actions the last

one might be a bit obvious but iterate

you're not going to get it right the

first time right if you ever put in chat

gbt type a question o that wasn't what I

wanted let me try it again iterate on

top of it you're going to have to try it

check it bounce it off users in the

large language model they did this this

term rhf reinforcement learning with

human feedback okay get that feedback so

in a nutshell get there fast enable it

broadly do it novel go to the actions

and iterate now let's think back again

right you're here at SMU and you're

thinking about what is this for me you

got have unique advantages you're young

you're adaptable you can go in the

workforce if you adopt these

things you're going to be productive

you're going to have folks that and

they're a long time they're tired of

that right they don't want to learn the

next thing now the other thing is what

if you want to do the entrepreneur thing

there might never be a better time this

thing is so disruptive right touching

everything anyway think about that we'd

love to have you at AT&T join our data

science scholar program or talk to us

afterwards

[Music]