2025 AI : 10 Things Coming In 2025 (A.I In 2025 Major Predictions)
Summary
TLDRThis video script outlines 10 major AI advancements expected in 2024, highlighting the rise of AI agents capable of autonomous task completion across various fields like customer service and healthcare. It also discusses the evolution of voice assistants, true multimodality in AI systems, personalized AI models, enhanced reasoning capabilities, custom-trained models like those for legal research, and AI's role in healthcare. The script further explores autonomous vehicles, improvements in customer service AI, and the exciting developments in humanoid robotics, signaling a future where AI becomes more integrated and sophisticated in daily life and industry applications.
Takeaways
- ๐งโ๐ผ Agents are emerging as pivotal AI systems capable of autonomous task completion, with Google and OpenAI leading the way in developing specialized agents for various tasks.
- ๐ฃ๏ธ Voice assistants are advancing rapidly, with OpenAI and Google set to release more sophisticated and expressive models that can interact more naturally with users.
- ๐จ True multimodality is on the horizon, allowing AI to process and generate a wide range of inputs and outputs, including text, images, audio, and video, enhancing user interaction and experience.
- ๐ค Personalization features for AI models are in development, aiming to provide more customized and context-aware responses by leveraging user-specific data and preferences.
- ๐ง Improvements in reasoning capabilities and reliability are expected for future AI models, reducing errors and enhancing the models' understanding and responses to complex queries.
- ๐ Custom-trained models are being developed for specific domains like law, integrating vast amounts of domain-specific data to handle complex tasks with higher accuracy and reliability.
- ๐ฅ Healthcare and AI are converging, with companies like Google and Apple working on AI-powered health coaching services that provide personalized health recommendations based on individual data.
- ๐ Autonomous vehicles are expanding their presence, with companies like Waymo offering public rides in select cities, signaling a significant step towards driverless transportation.
- ๐โโ๏ธ Customer service AI is transforming the industry by offering real-time, customized support and facilitating transactions more efficiently than traditional methods.
- ๐ค Advancements in robotics are leading to more capable humanoid robots, with new models like Atlas showcasing enhanced dexterity and autonomy, indicating a future where robots perform a wider array of tasks.
- ๐ The cost of humanoid robots is decreasing, making them more accessible for research and development, potentially accelerating innovation and practical applications in various sectors.
Q & A
What are the 10 major things in AI that the video discusses for the year 2024?
-The video discusses the emergence of AI agents, advancements in voice assistance, true multimodality, personalization features, reasoning capabilities, custom trained models for specific use cases like law, AI in healthcare, autonomous expansion, customer service AI, and improvements in robotics.
What is the significance of AI agents as mentioned in the video?
-AI agents are pivotal as they represent a shift in AI capabilities, allowing systems to autonomously accomplish tasks set by users, impacting various sectors and potentially changing how we interact with AI systems.
How does the video describe the evolution of voice assistants in 2024?
-The video suggests that voice assistants will become more natural and expressive, with the ability to understand and respond to complex commands, and will be integrated into more devices for hands-free interaction.
What does true multimodality in AI entail according to the script?
-True multimodality refers to AI systems' ability to process and generate various types of input and output, such as text, images, audio, and video, creating a more comprehensive and interactive user experience.
How will personalization features enhance AI models as per the video?
-Personalization will allow AI models to remember user preferences and context, providing more tailored and relevant responses, thus improving the usability and user experience.
What improvements in reasoning capabilities does the video anticipate for AI models?
-The video anticipates that future AI models will have enhanced reasoning abilities, leading to more accurate and reliable responses, and a reduction in the occurrence of mistakes.
Can you explain the custom trained models mentioned in the video for the legal field?
-Custom trained models for the legal field, as exemplified by the partnership between Harvey and OpenAI, involve integrating extensive legal data and domain-specific knowledge into AI models to handle complex legal queries and tasks with high reliability.
How is AI being integrated into healthcare as discussed in the video?
-AI is being integrated into healthcare through specialized models like Google's Med Gemini, which can analyze complex medical data and provide personalized health coaching and recommendations through mobile apps.
What advancements in autonomous vehicles does the video highlight?
-The video highlights the expansion of autonomous ride-hailing services in major cities and the potential for self-driving vehicles to become commonplace, improving transportation and reducing the need for human drivers.
How will customer service AI change the customer service experience?
-Customer service AI will enable faster, more efficient, and personalized customer interactions, handling transactions, and providing real-time information, potentially reducing wait times and improving customer satisfaction.
What is the potential impact of advancements in robotics as mentioned in the video?
-The advancements in robotics, including more capable humanoid robots and those that can perform tasks autonomously without teleoperation, suggest a future where robots can undertake a wider range of tasks, from household chores to complex industrial processes.
Outlines
๐งโ๐ผ AI Agents Revolutionizing Task Automation
The video discusses the emergence of AI agents in 2024, emphasizing their ability to perform autonomous tasks assigned by users. Google's demonstration of various agent types like customer, employee, creative, data, code, and security agents is highlighted, showcasing the potential impact on how we interact with AI. While the technology is still in early stages, the development by major companies like Google and OpenAI suggests a significant shift in AI utility and capability.
๐ค Advanced Voice Assistants and Their Future
This section explores the advancements in voice assistants, demonstrating their potential through interactive demos. The new generation of voice assistants is shown to be far superior to predecessors, with capabilities like storytelling, emotional expression, and even singing. The market potential is discussed, with Amazon's Alexa and Google's voice assistant projects indicating a move towards more natural and hands-free AI interactions.
๐ True Multimodality in AI Systems
The script delves into the concept of true multimodality, where AI systems can handle various types of input and output seamlessly. NVIDIA's vision and Google's AI studio are highlighted as examples of this multimodal approach, which includes capabilities like visual analysis, poster creation, 3D object synthesis, and lecture summarization. The potential for personalized AI models is also mentioned, suggesting a more intuitive and customized user experience.
๐ง Enhancing AI Reasoning and Reliability
The need for improving AI reasoning capabilities and reliability is underscored, with references to comments by industry experts like Sam Altman. The discussion points out the limitations of current models like GPT-4 and the importance of developing smarter AI that can understand context and provide accurate responses consistently. The potential for AI to make fewer mistakes and provide more reliable information is a key focus.
๐ฅ AI in Healthcare: Specialized Models and Applications
Healthcare is positioned as a significant beneficiary of AI advancements, with specialized AI models like Google's Med Gemini surpassing GPT-4 in medical QA benchmarks. The development of AI models fine-tuned for medical applications is highlighted, along with initiatives by Google and Fitbit to provide personalized health coaching through AI. Apple's plans for an AI-powered health coach in iOS 17 are also mentioned, indicating a trend towards AI-enhanced personal health management.
๐ Autonomous Vehicles Expanding Transportation Options
The script discusses the real-world application of autonomous vehicles, with services like Waymo offering rides in select US cities. Tweets from users experiencing self-driving taxis are shared, highlighting the comfort and reliability of the technology. The Automated Vehicles Act in the UK is cited as a sign of expanding legal support for autonomous transportation, suggesting a near-future where self-driving cars become commonplace.
๐ค Next-Generation Customer Service AI
The evolution of customer service AI is explored, with Google's AI capabilities being utilized by Symbol Fashions to provide real-time, customized support. The script describes an interaction where a customer service AI agent assists a customer with a purchase, applying discounts, suggesting additional items, and confirming the order efficiently. The potential for AI to transform customer service by handling repetitive tasks and providing instant information is highlighted.
๐ค Advancements in Humanoid Robotics
The final section of the script focuses on the progress in humanoid robotics, with Boston Dynamics' new Atlas robot hinting at significant advancements in the field. The script mentions the capabilities of existing robots and the potential for robots to perform a wider range of tasks autonomously. The decreasing cost of robots and China's investment in humanoid AI are also discussed, suggesting a future where robots are more accessible and integrated into various aspects of life and work.
Mindmap
Keywords
๐กArtificial Intelligence (AI)
๐กAgents
๐กVoice Assistants
๐กMultimodality
๐กPersonalization
๐กReasoning Capabilities
๐กCustom-trained Models
๐กHealthcare and AI
๐กAutonomous Expansion
๐กCustomer Service AI
๐กRobotics
Highlights
Introduction to the 10 major advancements in AI for 2024.
Emergence of AI agents capable of autonomous task completion.
Google's demonstration of diverse AI agents including customer, employee, creative, data, code, and security agents.
OpenAI's development of agent software to automate complex tasks by taking over user devices.
Potential for AI agents to transform how we interact with and utilize AI systems.
Voice assistance advancements with more natural and expressive interactions.
Google's Project Astra showcasing the future of AI assistance with voice.
Amazon's plans to overhaul Alexa with integrated large language models and generative AI.
True multimodality enabling AI systems to process and output various types of data.
GPT 4.0's multimodal capabilities showcased through different features and applications.
Personalization features of AI models to enhance usability and user experience.
Importance of reasoning capabilities and reliability for future AI models.
Custom trained models for specific use cases like legal case law research.
Healthcare and AI advancements with models like Google's Med Gemini for medical QA.
Autonomous vehicles expanding into major cities and the impact on transportation.
Customer service AI revolutionizing business interactions and support.
Developments in humanoid robotics with advanced capabilities and decreasing costs.
China's investment in humanoid robotics aiming for global leadership in AI and robotics.
Concluding remarks on the upcoming AI and robotics trends to watch in the next year.
Transcripts
so in today's video I'm going to be
telling you guys a look at the 10 major
things that are going to be coming in
2024
sl22 for artificial intelligence I'm not
sure how long this video is going to be
but there will be time stamps so you can
take a look at everything that is going
to be released and the future is most
certainly exciting so let's not waste
any more time so one of the first things
that is basically here is of course
agents now the video you're currently
seeing behind you is a video that I
recently made on Google Go's agents now
this is why the you know the text right
here isn't I guess you could say synced
up is because what Google did and what
many other companies are trying to Pivot
to in the late 2024 and in 2025 is of
course agents now agents essentially
just refers to I guess you could say AI
systems that are able to I guess you
could say kind of do whatever they want
and go out and accomplish a kind of task
that the person has set out for them and
I think this is going to be one of of
the most pivotal moments that does
happen in AI because this is going to be
one of the moments that actually impacts
you now the reason this is going to
actually impact us and the reason you're
going to truly feel this is because as
you can see in Google's demo if I pause
it right here you can see that they're
trying to basically hit every single
agent aspect that there pretty much is
going to be in Google since they've
decided to I guess you could say kind of
call the red flag you can see right here
that they've got customer agents
employee agents creative agents data
agents code agents and security agents
and this video there's basically quite a
lot of different features it's a
50-minute video where I've kind of cut
the long drawn out Google I guess you
could say announcement into a shorter
video about where they talk about you
know employee agents and this is going
to be really gamechanging because it's a
complete difference in how we currently
use AI systems currently we use AI
systems in this in the sense in the
general sense that you know we talk to
these AI systems and we interface with
them in a chat interface and with this
one we're going to Simply ask the AI
system to do something outline some
parameters and then the AI system is
going to be able to go ahead and
accomplish those tasks and that is why I
think this is going to be probably one
of the most pivotal moments now this is
still pretty pretty early in its stages
like currently we're still in a very I
guess you could say basic area for the
agents and the current demos that I'm
seeing from some of the top labs they
don't seem too too impressive so it
seems like there is a breakthrough that
is you know probably needed but I think
whenever this breakthrough does happen
because some of the larger larger
companies are really really working on
this I'm pretty sure this is going to be
crazy now one of the things is that open
ey have actually spoken about this you
can see that open ey have shifted its
Battleground to software that operates
devices and that automates t tasks so
essentially you can see right here it
says open ey is developing a form of
agent software to automate complex tasks
by effectively taking over a customer's
device the customer could then ask the
chat GPT agent to transfer data from a
document to a spreadsheet for analysis
for instance or to just completely do
random different things or to fill out a
spreadsheet and I think like I said
already this is going to be one of the
main things that really really changes
the game and as you know open AI have a
very very long history of releasing very
good products so whatever agent they're
working on apparently they're working on
two agents you know computer using
agents and these agents can effectively
take over a user's device to perform
tasks such as transferring data from a
document to a spreadsheet filling out
expense reports and entering them into
accounting software and these tasks
involve actions like you know clicks
cursor movements and text typing
mimicking basically human interactions
with different applications and of
course secondly other than the computer
agents okay and this is basically just
summarized from this article we have
web-based agents and these web-based
agents are going to handle tasks that
are primarily conducted online and they
can gather public data about companies
tra travel itineraries win of the budget
book flight tickets they can utilize
llms um and perform those really cool
tasks so basically what we have here is
I guess you could say a really really
really interesting bit now Sam Alman did
actually talk about this on about how
agents are going to be pretty pretty
incredible this is from an interview
that he did earlier this year with Bill
Gates it was one that I guess a lot of
people did Miss because Bill Gates
doesn't really have a large YouTube
presence but it was one that I managed
to get a small small audio snippet from
where Sam Alman actually speaks about
the future of agents and exactly what is
going to happen creely new things and
we're we're really seeing that and so as
we look as we look at these back steps
of things that can do a more complete
task you can like imagine a little agent
that you can say go write this whole
program for me I'll ask you a few
questions along the way but it won't
just be writing a few functions at a
time that'll enable a bunch of new stuff
uh and then again it'll do even more
complex stuff someday maybe there's an
AI where you can say you know go start
and run this company for me and then
someday there's maybe an AI where you
can say like go discover new physics and
it's the stuff that we're seeing now
very exciting and wonder
uh but I think it's worth always putting
in context of this technology that at
least for the next 5 or 10 years we'll
be on a very steep Improvement curve um
these are the stupidest the models will
ever be but coding is probably the area
the single area from a productivity gain
we're most excited about today uh
massively deployed and you know it's
scaled usage at this point Healthcare
and education are two things that are
coming up that curve so you could hear
that he was basically saying that it's
going to be a steep Improvement for
agents because it's of course something
that they haven't really worked on but I
guess agents are going to be a thing
where you can say hey look I need you to
go create this company for me I need you
to go run this business for me I need
you to go make this PowerPoint
presentation go do all of my homework or
whatever of course at that time I
guarantee there's going to be a
different kind of homework but the point
here is that you can see that agents are
truly going to be that breakthrough
moment and I do think that a lot of
companies are working on this now I
think the Breakthrough is likely to come
and potentially you know maybe late 2024
to early 2025 but I do think that agents
is something that you know Google as
you've seen are working on an open AI
are working on because it's very very
very important now in addition Google uh
have actually stated that this is what
they're working on and this is a clip of
demes hassabis the leader of Google's
Deep Mind which is their AI Division and
he speaks about how they're going to
develop these kinds of Agents I'm also
really excited about the next stage of
these large General models um you know I
think the next things we're going to see
perhaps this year perhaps next year is
more agent-like Behavior so systems that
are able not only to just you know
answer questions for you but actually um
plan and act in the world and and solve
goals you know and I think those are the
things um that will make these systems
sort of the next level of usefulness in
terms of being uh a useful everyday
assistant and again that's bringing back
in some of our uh work we did years ago
with with gaming which were all agent
systems um you know achieving certain
goals and objectives uh bringing that
kind of work and marrying that together
with the modern large multimodal model
work so yeah there's Deus saus talking
about the future of agents and I think
he's very much to right in the fact that
that is going to bring things to a
completely another level um and it's
going to be completely different because
this is where you're going to unlock
real real value because a lot of people
um I wouldn't say they don't use AI
systems but it's going to be interesting
to see how you know using those AI
systems is you know vastly different
than the way we interact with them now
so I think that is going to be super
super fascinating now here we are for
the next part okay and the next part of
course is voice assistance now many of
you may have seen the new open AI demo
and this is of course what we're going
to be seeing a lot more of from the top
Frontier AI labs in 2024 and of course
early 20125 now there's actually a lot
to discuss here because Voice Assistant
might not be what you guys think so
essentially in this demo here we
actually did get to see a huge huge huge
huge variety of different voice
assistants and these voice assistants
were really really really good because
compared to the previous ones that we've
had before like Siri they completely
brow blow just blow everything out of
the water so this is a short gbt 40 demo
and then I've got to show you guys
another demo that most people did
actually miss from another top Frontier
company that I guess you could say is
working on something very very similar
to this hey Chach PT how are you doing
I'm doing fantastic thanks for asking
how about you pretty good what's up so
my friend um Barrett here he's been
having trouble sleeping lately and uh I
want you to tell him a bedtime story
about robots and love oh a bedtime story
about robots and love I got you covered
gather round Barrett once upon a time in
a world not too different from ours
there was a robot named B
bite was a curious robot always expl
started this story but I want a little
bit more emotion in your voice a little
bit more drama got it let's add some
drama once upon a time in a world not
too different from ours there was a
robot named I really want maximal
emotion like maximal expressiveness much
more than you were doing before
understood let's amplify the drama once
upon a time in a world not too different
from ours there was a robot named bite
bite can you do this in a robotic voice
now initiating dramatic robotic voice
once upon a time in a world not too
different from ours there was a robot
named bite bite was a curious robot
always exploring new circuit Barrett
here calm down a little bit can you end
the story um but do it in a singing
voice and so bite fell owned another
robot friend and they live circly Ever
After thank you so much I definitely
feel a lot more
relaxed so now the reason voice
assistants are going to be pretty pretty
crazy and the reason like a lot more
people are going to be using them is
because we've already seen the fact that
the The Voice assistance are serving a
pretty huge Market if you remember the
rabbit R1 device that was I guess you
could say a Voice Assistant that was
much better than anything else at the
time and although many people are
stating that it's just a rapper product
or yada y yada I think the point here is
that voice assistants are truly the way
that we are going to interact with AI
because it just feels the most natural
when I'm talking to GPT 40 or when I'm
talking to you know another person I'm
really trying to feel the same way and
the same kind of fluidity and that's
what most people use of course there are
going to be some scarios where you just
simply want to you know talk with a
large language model or an AI system and
get out you know uh lots of lots of text
but I think that for the vast majority
of people on a day-to-day basis you know
handsfree being able to just talk with
your AI about whatever topic it may be
this is going to be the large majority
of where people are you know Finding
real real use for these systems okay uh
and this also is great because for those
who aren't technologically advanced uh I
guess you could say this is going to be
a much easier way for them to interact
with the software now another thing that
we do have is Google's Project Astra so
this is the vision for Google's future
of AI assistance and this is quite
similar to the recent thing that we just
saw but it's Google's very own
interesting way of doing their voice
assistance and this is likely to be
released in the coming months although
Google haven't really shipped anything
that they like that they said they were
going to ship but this is of course
something thing that they're still
working on and this demo is giving us an
insight to where things are going to
head so this should be giving you guys
some creative ideas for how the
interaction and the ecosystem around AIS
is going to change okay let's do some
tests tell me when you see something
that makes sound I see a speaker which
makes
sound what is that part of the speaker
called that is the Tweeter it produces
high frequency sounds
give me a creative alliteration about
these creative crayons color cheerfully
they certainly craft colorful
Creations what does that part of the
code do this code defines encryption and
decryption functions it seems to use a
ecbc encryption to encode and decode
data based on a key and an
initialization Vector IV that's right
what neighborhood do you think I'm
in this appears to be the king's cross
area of London it is known for its
railway station and transportation
connections do you remember where you
saw my glasses yes I do your glasses
were on the desk near a red
apple what can I add here to make this
system faster adding a cache between the
server and database could improve speed
what does this remind you of shringer
cat all right uh give me a band name for
this Duo golden Stripes nice thanks
Gemini so yeah that is Google's aim at
the voice assistants and I think voice
assistants have another area in which
they're going to be you know moving
forward because there are other
companies that have had a market share
and let me just put something on the
screen and that is of course Amazon so
Amazon of course are planning to give
Alexa an overhaul and a monthly
subscription price so Amazon is making
significant upgrades to Alexa by
integrating large language models and
generative AI capabilities Amazon is
actually developing a new version of
Alexa powered by custombuilt llms
optimized for voice interactions and
this llm whichever it's going to be
based on I think it most likely will be
claw 3 are going to enable more natural
con conv ation interactions with Electra
and some key enhancements include
conversation a visual wakeup word
conversational speech recognition and
expressive text to speech so this is
going to be another step in terms of
integrating these AI systems with the
wider wider world and we know that there
are already tons of people who already
have these devices in their homes that
they could literally just receive the
software update and then boom you now
have millions of users who are going to
be interacting with voice assistants
that are actually really really useful
so it's going to be interesting to see
what Amazon does and I'm sure that their
partnership with anthropic is going to
be a major part of that now like I said
it before one of the things that I do
think would be surprising is if Google
didn't make an announcement regarding
their home devices I'm not sure of the
popularity of these I know that most
people do have an Alexa I've never
really heard anyone talking about their
Google device but I'm pretty sure it
would sell a lot more devices if you
could just have a home device or a
physical device where you could just ask
it simple questions and it could respond
very easily I know that's something that
I would certainly be using so the next
thing that is on the list for AI number
three is of course true
multimodality so essentially any to any
and this is a screenshot from nvidia's
recent conference but it gives us a
glimpse into the future of how
multimodal systems are truly going to
work any kind of input that you can
think of is going to be used and then of
course you're going to be able to put
that into any other kind of I guess you
could say input/output and this
multimodality this true multimodality is
in some ways already here and will be
further expanded upon so one of the ways
that most people didn't even realize and
this is something that I actually saw
because I was paying attention and most
people didn't realize but that when GPT
40 was released they released this web
page okay and there were an exploration
of different capabilities and in the
capabilities section there were a bunch
of different you know things that showed
us that gbt 40 was and is a completely
multi modal model that's why they called
it GPT 40 Omni and there were many
different things that they just didn't
Showcase in the demo that proved to us
that this is the future of AI because
it's actually a completely multimodal
model natively built it's not like it's
a you know large language model or
something like that it's completely a
natively multimotor model so I'm going
to show you guys quickly what I mean by
this so if you go onto the gbt 40 web
page and if you scroll down there's this
section right here this section is
called an exploration of capabilities
you can see that this is visual then you
can see that this is poster creation and
you can see it's able to combine
different things you can see we have
here this photo character you can see
turning photos into different things
things you can see concrete poetry
you're able to turn this into words and
I think you're also able to turn this
into a badge you're also able to see
lecture summarization so this shows you
that you've got a 45 minute video and
it's able to take the input from that so
remember this is true true multimodality
meeting notes with multiple speakers
it's able to take in audio you can see
right there there was also 3D objects
synthesis so you can see right here it's
able to generate a 3D object and able to
rotate that in Real Time 3D
reconstruction so this is something that
is truly truly going to change
everything because we have a situation
where there's going to be true
multimodality which means the only
limits that you're truly going to be
thinking about are ones in your head and
in addition to that the multimodality
aspect has been there with Google's
recent AI studio so if you don't know
Google have released their AI studio and
in there you can actually upload audio
you can upload images and you can upload
video videos up to an hour long so it's
pretty pretty crazy what you can truly
do with true multimodality because
currently with Google's new studio it's
not out for everyone although it's out
for most people but still when I'm
trying to use it it's still pretty buggy
so it's not something that I've actually
been able to use into my workflow just
yet but true multimodality I do think is
going to change a lot of things because
whilst images and text are good videos
and audio are going to complete the
entire equation then what we have is
very interesting we have the
personalization of these models so
basically there is a feature that is
only available to a select few accounts
in select few regions namely USA being
one of them and that is of course the
personalization feature so some in the
later months I'm sure that
personalization is going to be rolled
out to Europe and the rest of the UK and
I'm sure that when it does it's going to
increase the I guess you could say the
usability of these models because
personalization is something that
changes the game you don't need to
remind it all the things you've done
before you don't need to remind it all
of your preferences and it completely
knows all the facts that you want it to
know about you that make it a lot more
easier to use on a day-to-day basis like
for example I was trying to ask it a
question okay and I was like okay what
is the best restaurant or what type of
food should I eat today and you have to
remember that when you open a new chat
with chat GPT or any AI system it
doesn't have this area of memory which
means its only data point that it has is
that a human wants to eat some kind of
food it doesn't have your age it doesn't
have your location it doesn't know what
you like to eat it doesn't know if
you're allergic to anything all it knows
is that human wants to eat food so when
you give it that command it's going to
be able to give you a very broad answer
that's very vague that can satisfy all
of those things however if you're like
okay I'm a male I go to the gym I'm very
active I'm trying to get Le C it might
suggest here some protein especially if
you like a sweet taste try this with
honey like some beef and like it will
try to suggest things that actually are
useful for you and that's going to give
a huge huge huge boost in terms of the
usability because the more data you give
these models the the better they are and
that's something that I've known in my
personal experiences when I can't get
these models to do something right I
just give it more and more context and
the more context I give it the better
the responses from these models are and
here you can see Sam Alman and both
Gates actually speaking about this
customizability and personalization will
also be very important people want a
very different very different things out
of gp4 different styles you know
different sets of assumptions we'll make
all that possible and then also the
ability to have it use your own data so
the ability to know about you your email
your calendar how you like appointments
booked connected to other outside data
sources all of that those will be some
of the most important area so yeah
that's um Alman talking about how those
are some of the you know most important
areas for the AI systems um and of
course the next thing that we do have
and this is actually Point number five
although it does say number four is of
course the reasoning capabilities um and
the reliability of the model that just
means the model doesn't make mistakes
and the model is a lot smarter we'll be
able to push that much further but maybe
the most important areas of progress
will be around reasoning ability right
now gp4 can reason and only extremely
limited ways and also reliability you
know if you if you ask gp4 most
questions 10,000 times one of those
10,000 is probably pretty good but it
doesn't always know which one and you'd
like to get the best response of 10,000
each time so that'll be that that that
increase in reliability will be
important and so yeah Sam mman says
reliability and reasoning ability is
going to be pretty pretty important now
samman has spoken about this and if you
don't know what he means by that he
basically just means that the model is
just going to really understand truly
what you want and it's never going to
make a mistake again I wouldn't say
never because I think that's pretty hard
to say but um you know it's pretty much
going to be not as much as it does right
now so he talks about this and I've
included this clip here because he
basically says how gp4 is quite stupid
and it's quite dumb uh and that's to
point out the reasoning capabilities of
the model are pretty limited and if you
don't know what I mean if you want to
take a quick demo you can like there was
a question that people used to ask to
GPT 4 and the question that they used to
ask it was like if someone moved
something from one bag to another but
the bags were transparent um could you
see what's inside and the model doesn't
realize that the bags are transparent
and that's just basically uh part of its
reasoning capability to be like wait a
minute the bags are transparent I should
be able to see exactly what's inside um
but it can't so things like that you
know you you might not use it on a
day-to-day basis if you're using it to
write essays but um an increased
reasoning capability means that there's
going to be a lot more wide applications
which means it's going to impact a lot
more industry so uh here's where samman
talks about the fact that gpc4 is quite
dumb um I I mean this this sounds like a
copout answer but I think the most
important thing about GPT 5 or whatever
we call that is just that it's going to
be smarter and this sounds like a Dodge
but I think that's like among the most
remarkable facts in human history that
we can just do something and we can say
right now with a high degree of
scientific certainty GPT 5 is going to
be smarter than a lot smarter than GPT 4
GPT 6 is going to be a lot smarter than
GPT 5 and we are not near the top of
this curve and we kind of know what know
what to do and this is not like it's
going to get better in one area this is
not like we're going to you know it's
not that it's always going to get better
at this eval or this subject or this
modality it's just going to be smarter
in the general sense and I think the
gravity of that statement is still like
underrated like I said before the
gravity is definitely going to be
underrated because of the vastly smarter
model we've seen what the capabilities
are the jump from GPT 3.5 to4 meant that
people could actually use GPT 4 for you
know certain things like code certain
things like writing uh you know a lot
more I guess you could say things that
you know a lot more meaningful um and
there are just a lot more tasks that you
can do and of course this is where he
talks about the fact that gbt phenomenal
chat gbt is phenomenal um everything
else all the other models are phenomenal
it burnt you've earned $520 million of
cash last year that doesn't concern you
in terms of thinking about the economic
model of how do you actually where's
going to be the monetization source well
first of all that's nice of you to say
but Chachi PT is not phenomenal like
chbt is mildly embarrassing at best um
gp4 is the dumbest model any of you will
ever ever have to use again by a lot um
but you know it's like important to ship
early and often so yeah pay attention to
what he said there it's going to be the
dumbest model you have to use for a
while by a lot so I think that's going
to be something that's really really
really important now coming in at number
six is of course custom trained model so
here what you can see is a uh model that
is trained for Law and you can see that
it's going against gp4 so basically
Harvey the company Harvey developed a
partnership with open ey and they
developed a custom case law model so
basically they have foundational models
and they use public apis and they
attempted to fine-tune Foundation models
using public apis and retrieval
augmented generation systems but these
systems weren't really good because they
needed deep domain knowledge and
sophisticated reasoning and then they
decided to collaborate and partner with
open AI to train a model specifically
for legal case law research and this
involved integrating a vast amount of
legal data and domain specific knowledge
into the base models and then they added
case law from Delaware and gradually
expanded to include the entire Corpus of
the US case law and this added around 10
billion tokens worth of legal data to
the model so this was a highly
collaborative effort from open AI with
lawyers explaining the intri intricacies
of case law research and AI researchers
demonstrating model capabilities and
this is where openi provided insights
into the technical aspects and
adjustments needed to refine the model
so the custom train model can handle
tasks that require detailed legal
reasoning and the ability to draw from
extensive legal knowledge and it's
capable of generating comprehensive
answers to complex legal queries and
drafting documents and one of the main
improvements is that this model's
ability to reduce hallucinations and
ensure that every statement is backed by
cited case law and this makes the
outputs from this model highly reliable
and useful for legal professionals who
need accurate references so overall what
we can see here is the fact that in the
future we're going to have models that
are custom trained for specific use
cases um and law is going to be one of
them whilst yes GPT 4 can be used to you
know argue certain things I think it
shows here that when you have a custom
trained model the use cases go up a
significant amount um and that is
something that we're probably going to
see a lot more more often now next is
something that I think we are definitely
likely to see because the rate of
improvement here has been very very
interesting so this is Healthcare and Ai
and there are a few companies that have
been silently working behind the scenes
to I guess you could say develop certain
Healthcare products that are going to be
ready very very soon and I think they're
going to take advantage of these new AI
systems and it's going to be a really
really interesting thing currently what
your looking at is of course the
benchmarks on the med QA the medical
question and answering QA and you can
see right here that GPT 4 has been
surpassed by Google Geminis Med Gemini A
specialized version for answering
medical questions so essentially it's a
family of AI models fine-tuned
specifically for multimodal medical
applications and those models are
capable of understanding and analyzing
complex Medical Data including text
images videos and electronic health
records like I said the multimodal
capabilities of Gemini mean that these
models are now able to digest a variety
of different information and do Advanced
reasoning on it and it's shown
remarkable performance achieving
state-of-the-art accuracy of
91.1% on the med QA Benchmark which is
of course used for evaluating medical
knowledge similar to the US medical
licensing exam and like I said it's not
just research Google research and Fitbit
are actually building personal health
llms to offer personalized coaching and
recommendations through a mobile app so
leveraging Google's family of generative
AI models in Gemini and sufficient
Health Data from Fitbit and pixel
devices this initiative aims to provide
tailored guidance based on individual
health and fitness objectives so Google
have just recently announced that they
are working on this and they are
actively trying to bring this to Fitbit
and not just Google but apple apple
plans a AI powered health coach coaching
service a mood tracker and iPad Health
app so this is the move part of iPhones
make us push into health services so
this is going to be something that is
planned for iOS 17 maybe we're going to
get an announcement at Apple's new not
conference but I think it will be
interesting to see what actually happens
because this is something that I think
is very very underrated and this is
definitely going to impact Healthcare in
a very positive way because it's going
to bring down I guess you could say
access to high quality information next
of course we have something that is the
autonomous expansion so if you don't
know autonomous cars are a real thing
and they're basically cars that drive
without you they don't need a driver and
you can sit in the passenger seat and
you know get taken to your destination
it's really weird but it actually makes
a lot of sense so essentially what
you're looking at here is a few tweets
from people that have you know been
firsthand in these Whos so you can see
right here this person says so weo is
actually amazing using it in SF for
every trip now and it's just rock solid
on a ride now and it just navigated a
bunch of really odd and expected on
Street nonsense with confidence and ease
and someone says I now prefer
self-driving Whos over Ubers I actually
feel safer and someone said I just tried
Whos self-driving taxis it's more
comfortable than taking an Uber AI is
changing everything even for taxi and
Uber and the thing is is that this is
expanding this kind of stuff is going to
be expanding into many different areas
and eventually I do think that it's
probably going to be across the US but
you can see here that it says later this
year weo will be offering rides to the
public in four major cities starting
tomorrow March 14th our fully autonomous
ride hailing service will be available
to select members of the public in Los
Angeles and after starting initial Rider
only testing in Austin last week we plan
to offer wh one to austinites later this
year so that is pretty crazy they going
to be expanding this and we're likely to
see a lot more of the user feedback from
this and I think with some of the other
announcements especially going on in the
UK as well with the fact that the
automated Vehicles Act is now law it
means that self-driving Vehicles could
be on British roads in Just 2 years and
it's going to be a very very interesting
time now with that we have Point number
nine now what else is coming and this is
customer service AI so if you're someone
that interacts with the business which
is all of us anytime you have to speak
to customer service there is usually
some tedious interaction but that's
about to change take a look at Google's
customer service AI That's changing the
game behind the scenes symbol Fashions
customer agent is using Google Cloud's
full Suite of AI capabilities to offer
customized support interactions you know
facilitate transactions like purchases
and returns and ensure that I'm
receiving the most upto-date information
in real time I'm so close to having this
shirt for the concert let's give the
store a call
hi there this is the symbol fashion
customer agent at South Las Vegas
Boulevard am I speaking with Amanda yes
this is Amanda great thanks for reaching
out Amanda I see you had a session on
another device I've sent you an SMS
message with a link to our live chat
companion if you would like to switch to
chat please click the link how can I
help you today I'd like to purchase the
shirt in my cart with the cart I have on
file absolutely I see you're also a
symbol fashion Awards member looks like
you have a 20% off voucher available to
use would you like to apply it to this
purchase yes please that would be great
the shirt you're purchasing goes well
with these items also available for
pickup in your preferred size would any
of these be interesting to you
absolutely please add the white shirt
and the boots to my cart great your
total is $
23.73 okay to proceed with the card on
file yes your purchase is confirmed do
you need anything else today no I'm all
set thank you incredible thank you for
shopping with symbol fashion you'll get
a text when the items are ready for
you in less than 5 minutes I was able to
find and order the shirt I wanted and an
outfit to match I'm ready for the
concert back to you Thomas many of so
you can see right there that that is
truly going to change the game and and I
think it's I think it's real and in the
sense that like this is going to be
something that actually helps companies
out because you know customer service is
just tedious you got people always
complaining people not being able to
find stuff and I think ai's ability to
you know uh rapidly access information
be able to be patient you know not get
annoyed if someone raises their voice I
think this is going to be an area that
humans don't even want to do anyways so
this is going to be something that's
really impressive now of course with
customer service AI you do have other
areas here uh and Wendy's has been
actually I guess you could say testing
this out i' I've shown this a couple of
times but this is very relevant because
I think that you know more
drive-throughs are probably going to use
this stuff provided it's effective as as
it's shown
here welcome to Wendy's what would you
like can I have a chocolate
frosty which size for the chocolate
frosty
medium can I get you anything else today
no thank
you great please pull up to the next
window so yeah you can see uh the
differences in how things are going to
go and I think this would be really good
because I've seen too many videos of
people taking the Mickey out of people
in drive-throughs and it's quite
frustrating cuz some people just want to
work their jobs and people are like
trying to make funny content out of them
so if we just had AI in this position
that could speak all the languages
understand exactly what you said be
completely patient I think that that's
going to be something that's really
really cool now another thing that we do
have and coming in at Point number 10
last but not least is of course even
better robots so we currently looking at
is boss and Dynamics new Atlas which is
pretty pretty incredible we haven't seen
a full demo yet but from what they've
shown us they've shown us that they are
about to embark on a new humanoid
robotics run that if this robot can move
anything like their old Atlas I think
people are going to be in for a real
real surprise on how this robot is going
to move considering it's completely
electric um and I don't think it's
hydraulic based like the previous one so
we're about to get a huge new I guess
you could say not robotic spoon but like
a huge huge focus on robotics because
robots have been I guess just going
through an incredible period of growth
and you can see here that even recently
there was this robot here at one time
speed being able to do just incredible
tasks that I still struggle to believe
that this is a robot with no
teleoperation so it's incredible at what
some of the companies around the world
are being able to do with robotics so I
think you know later on this year or
even in 2025 there's going to be a few
things that show us that you know robots
are going to be doing a lot more
different tasks and this is especially
combined with the fact that you can see
right down here these things are
combined with some of the multimodal AIS
and of course these llms integrated with
them so it's going to be a very very
interesting time to see what kind of
things we do get with robots and not
just the fact that this is like one time
speed no teleoperation which is just
genuinely mind-blowing that a robot is
able to just do this because it now
means you know in the future when you
know people are talking about robots
doing a lot of household tasks and
things like this and remember guys this
is no teleoperation which means there's
no human uh in another room that's
controlling this this is a robot
completely autonomously doing this based
on a set task um and I think you guys
need to also take a look at the fact
that China has been developing a ton of
humanoid robots this is only like half
of them um and so far the demos that
we've seen the infrastructure that
they're they're building you know
everything is being you know put in
towards humanoid Robotics and Ai and I
think China definitely wants to be a
leader in that space because they
understand the economic value and even
recently we got unry G1 which is an AI
agent Avatar as described and it's a
humanoid robot that is only
$166,000 which means the decreasing cost
of these robots means that researchers
can afford to purchase these systems and
apply their research techniques on these
systems because a lot of times the the
problem is that you know these robots
are pretty pretty expensive especially
some of the cool ones that you're seeing
in these demos they're like $250,000 to
half a million dollars and I mean if
you're a researcher at a university it
can be pretty pretty impractical to just
buy one so that you can do some research
Arch on it but if we can get these
things down with economies of scale it
could open up a huge huge huge area for
where we're seeing rapid progress on
these devices so that was the 10 things
that I think are going to be coming in
the next 365 days for AI robotics slth
Singularity and those are going to be
things that you truly do see here on
this channel if there was anything I did
miss don't forget to leave a like down
below don't forget to comment down below
and let me know what thing you are most
excited for cuz I would love to know
Browse More Related Video
ChatGPT 5 Will Revolutionize Everything: Hereโs What You Must Know
GPT-5 SOON, AI-to-AI Payments Using Crypto, xAI GPU Cluster is Live, 1,000 Agent Simulation
The AI opportunity: Sequoia Capital's AI Ascent 2024 opening remarks
AI Unveiled beyond the buzz episode 4
AI News: The AI Arms Race is Getting INSANE
BIG AI NEWS: 10,000X Bigger Than GPT-4, AGI 2025, New Boston Dynamics Demo And More
5.0 / 5 (0 votes)