Google's NEW AI 'PALM-2 Takes the Industry By STORM! (NOW RELEASED!)

Key Takeaways

so Google once again has tried to up the bar because of course as you know the AI race continues Google the multinational company decided that what they want to do now is up the bar with palm 2. so essentially Palm 2 is going to completely change the game in terms of large language models but you should take a closer look at what they've announced and why it's going to change many different Industries so let's take a look so now we have Palm 2. now what are the defining features which makes Palm 2 so honestly creative and so cool is that it's actually built in multiple stages there are five different stages for Palm 2 which is essentially their large language model that's going to pretty much power every single application and essentially what's so cool about this is that they have even a lightweight version called gecko which is built to run natively offline on your phone which is essentially a large step in the large language models because this now means that essentially we're going to be getting large language models embedded into other applications which work offline which is truly honestly insane so this is something that I honestly can't wait for and palm 2 was trained on a bunch of different things and it's honestly improved quite a bit from the original palm which we already know was quite a success when embedded into some robots and it actually supports over a hundred different languages so it's going to be interesting to see what many different communities create then one of the major upgrades to Palm 2 was its ability to code we know that coding is something that many people use chat TPT for but with palm 2 they actually now support coding in many different languages and with the Palm 2 Integrations and upgrades essentially what we have is a more robust and sophisticated version that can even code in many different languages which is honestly quite impressive I'm not even sure if this is something that chat GPT can do at the moment so this is a major breakthrough for Google and of course we have Palm 2's fine tuning which is honestly insane because of course as you know large language models can be fine-tuned to perform very well at specific certain tasks so what Google actually did was they decided to fine-tune a version of palm which was essentially focused on medical research and diagnosis and essentially with palm 2 now you can get accurate medical reports and diagnosis on certain cases which is very very impressive because we knew that this kind of stuff is very very hard to diagnose correctly and recent studies showed that estimated errors May account for as many as 251 000 deaths annually so this is a big problem that Google are trying to fix and with the invention of medpom 2 there's a nine times reduction of inaccurate reasoning so essentially they are improving the accuracy results of these diagnoses that are given and it seems to be working very well so you can see right here that the llm performance before 202 G3 and then we have Med Palms 2 actual Benchmark reaching the expert performance level which is honestly quite impressive because this stuff is moving very very quickly and they've already reached X performance right here you can literally see that this is of course multimodal because it says what does this film show and then mid Palm 2 gives the right feedback so it shows what kind of fracture there is and this is honestly quite interesting because you can literally input a scan and then the AI is going to tell you exactly what is wrong which is going to literally solve a lot of issues when it does come to wrong diagnosis taking time to diagnose certain things and of course contrasting opinions on what is wrong so this is something that is really really good for the healthcare industry and I can't wait to see this integrated into many systems which is going to make a very healthier and safer world so essentially one thing that Google actually did earlier this year was around April the 4th 2023 they actually released a groundbreaking research in health AI where they actually did a deep dive into Palms 2's Med Palm now this kind of went under the radar with the video only getting around 16 000 views but this was a truly huge advancement for medpom because as you already know medpom has shown to have many advancements in AI now one of the key things that makes medpalm so great is that of course it performs exceedingly well compared to the previous benchmarks that we've previously seen now this is a true breakthrough because this was something that they didn't expect you see you can currently see that the previous benchmarks of previous large language models before 2023 or essentially before medpalm was released and you can see that they don't actually hit the required Benchmark in order to pass certain tests but of course as you know when they released the first version of medpalm which is based on Palm 2 you can see that they immediately got the 67.2 percent and of course in the future which the version which we have now is of course at currently around 87 what's interesting is that you can actually see clinicians answers compared to Med Palms answers and essentially what they did with medpom was they decided to rate these answers on a completely new rating system to see if medpom's answers were actually good now you can see that for each of these questions they decided to get real clinicians to answer the same questions as medpom and then what they then had was clinicians versus the medpalm to see what other clinicians thought the answer was in terms of if it was good and you can see right here the clinician review of medpom's answer shows that it was actually pretty accurate now this is of course really good because this definitely can Aid certain doctors when it comes to giving out diagnosis and it was previously discussed this is one of the biggest issues in healthcare because many times millions of people do die from wrong diagnosis we also have another example here where you can see that there is the clinician answer and the medpalm answer and I think in this example Google Health dictated that the medpom answer was arguably somewhat a little bit different but gave certain insights that the clinician's answer didn't give so if you're wondering if this med Palm is going to be completely replacing the healthcare industry I don't think so but I do think it is going to be a huge boost because these large language models have the ability to analyze large amount of data sets I remember in a recent AI podcast I believe it was Eliza yudowski that essentially talked about how if you would prefer to see a doctor that had seen maybe a hundred thousand patients rather than a doctor that had seen 100 million patients I'm pretty sure most people would be comfortable with seeing the doctor that had a hundred million patients okay and that's because of course AI can essentially see that many patients because all you need to do is train it on many different specific cases and once it gets really good at spotting certain things it's going to be essentially increasingly good at making accurate predictions which is going to be great because now you don't have that human error where people may make misjudgments people can make complaints and of course there can be errors in human you know analysis so this is going to be something that really does shake up the healthcare industry and for a good way not in a bad way at all because this won't actually replace anyone this is only going to assist those that are already going to be confident in their decisions if you're wondering just how powerful Palm is when it comes to converting different languages and translating them you can see right here this is an excerpt from the research paper that Google Google actually produced and it talks about how you can actually fix a bug with comments in Korean so you can see that the code is in English and then essentially what we have here is someone that said fix a bug and then add these comments in Korean which is actually really really good because this now means that this kind of software is going to be effective in many different countries now I'm not sure that gpt4 does have this capability but it does seem that Google is trying to excel at areas where gpt4 and other large language models do lack and I think this is somewhere that they've already shown that they are really really doing well in now another piece of information for Palm 2 was the ability to out beat Palm one now of course this is expected but these benchmarks show just how greater Palm 2 is compared to Palm one and I think there's a little bit of stuff in this research paper that they honestly don't talk about too much but are honestly really interesting for example there's also this from the paper you can see right here that they talk about how Palm compares to gpt4 and that it shows that it's actually quite competitive now you can see right here from this table you can see GPT 4's results on certain tests then you can also see Palm 2's results on certain tests and you can see that these large language models are starting to catch up to one another and palm 2 is actually quite competitive with gbt4 now now the only question I have for you is Palm 2 going to be ever catching up to gpd4 or is Palm 2 just going to be what gbt 3.5 is and like the other chat Bots like Claude instant so let me know what you think about Google's Palm 2 and the innovations that they've done and if you'll be using this or not