WEBVTT

1
00:00:00.440 --> 00:00:04.879
<v Speaker 1>Welcome to AI assembled the top AI news for small business.

2
00:00:05.360 --> 00:00:08.080
<v Speaker 1>I'm Chris daily and each week I bring you the

3
00:00:08.160 --> 00:00:13.519
<v Speaker 1>most important AI developments that can transform your business. Let's

4
00:00:13.560 --> 00:00:17.920
<v Speaker 1>dive into this week's top stories. There is a quiet

5
00:00:17.960 --> 00:00:21.160
<v Speaker 1>counter movement building against the handful of tech giants that

6
00:00:21.239 --> 00:00:26.120
<v Speaker 1>dominate AI to day. It is called Current Ai. It

7
00:00:26.199 --> 00:00:29.239
<v Speaker 1>is a non profit and its ambition is to build open,

8
00:00:29.359 --> 00:00:33.119
<v Speaker 1>public AI infrastructure the way the early World Wide Web

9
00:00:33.240 --> 00:00:36.719
<v Speaker 1>was built as a shared resource rather than a private

10
00:00:36.759 --> 00:00:41.159
<v Speaker 1>toll road. Current ai is backed by roughly four hundred

11
00:00:41.280 --> 00:00:45.200
<v Speaker 1>million dollars in initial funding. That money comes from the

12
00:00:45.200 --> 00:00:50.560
<v Speaker 1>French government, the Ford Foundation, the MacArthur Foundation, Google deep Mind,

13
00:00:50.920 --> 00:00:56.200
<v Speaker 1>and sales Force, among others. Last week, the organization launched

14
00:00:56.200 --> 00:01:00.759
<v Speaker 1>an open source AI chackbot called Alpha Chat, and last

15
00:01:00.840 --> 00:01:04.719
<v Speaker 1>month it deployed three point two million dollars in grants

16
00:01:04.760 --> 00:01:08.760
<v Speaker 1>to public interest AI projects in Kenya, Lebanon, and the

17
00:01:08.799 --> 00:01:14.599
<v Speaker 1>Brazilian Amazon. Those grants are funding local language models, cultural

18
00:01:14.640 --> 00:01:19.480
<v Speaker 1>preservation tools and AI built by and four communities that

19
00:01:19.560 --> 00:01:25.120
<v Speaker 1>the big frontier labs are frankly ignoring. Here is what happened,

20
00:01:25.480 --> 00:01:28.079
<v Speaker 1>and here is why it matters, and here is what

21
00:01:28.200 --> 00:01:28.439
<v Speaker 1>to do.

22
00:01:28.560 --> 00:01:32.920
<v Speaker 2>About it. What happened is that a serious, well funded

23
00:01:32.920 --> 00:01:37.640
<v Speaker 2>alternative to the closed, subscription based AI world is finally

24
00:01:37.680 --> 00:01:41.519
<v Speaker 2>getting off the ground. Why it matters is that if

25
00:01:41.519 --> 00:01:44.920
<v Speaker 2>the only AI your business can access is owned by

26
00:01:44.959 --> 00:01:49.000
<v Speaker 2>three or four American companies, you are one pricing change

27
00:01:49.200 --> 00:01:53.760
<v Speaker 2>or one policy change away from being cut off. Now

28
00:01:53.840 --> 00:01:57.079
<v Speaker 2>the important part, which is what a small or mid

29
00:01:57.159 --> 00:02:03.040
<v Speaker 2>sized business should actually do. Number one, put Alpha Chat

30
00:02:03.079 --> 00:02:06.079
<v Speaker 2>and the other open models current AI is funding on

31
00:02:06.159 --> 00:02:10.319
<v Speaker 2>your evaluation list this quarter. You do not have to switch.

32
00:02:11.039 --> 00:02:13.400
<v Speaker 2>You just need a working plan B for the day

33
00:02:13.520 --> 00:02:17.280
<v Speaker 2>open AI or Anthropic raises prices on you, and that

34
00:02:17.479 --> 00:02:22.560
<v Speaker 2>day is coming. Number two. If your business operates in

35
00:02:22.599 --> 00:02:26.360
<v Speaker 2>a non English market or serves a community whose language

36
00:02:26.360 --> 00:02:31.039
<v Speaker 2>is under represented, check the current AI grant list. Some

37
00:02:31.159 --> 00:02:33.919
<v Speaker 2>of the models being funded right now will speak your

38
00:02:33.919 --> 00:02:37.599
<v Speaker 2>customer's language better than g P T or claud ever

39
00:02:37.680 --> 00:02:42.120
<v Speaker 2>will Number three, if you have any interest in your

40
00:02:42.199 --> 00:02:46.479
<v Speaker 2>data staying on infrastructure, you can audit open source Public

41
00:02:46.560 --> 00:02:51.840
<v Speaker 2>interest AI is the only long term answer. Start experimenting

42
00:02:51.879 --> 00:02:55.400
<v Speaker 2>now at low stakes so you are not scrambling later.

43
00:02:56.639 --> 00:03:00.439
<v Speaker 2>Knowing this story exists is only the first step. The

44
00:03:00.560 --> 00:03:03.800
<v Speaker 2>move is to actually pilot one of these tools before

45
00:03:03.840 --> 00:03:08.319
<v Speaker 2>your competitors do. Now to a story that could quietly

46
00:03:08.400 --> 00:03:13.199
<v Speaker 2>reshape health care, insurance, and eventually what your employee benefits

47
00:03:13.199 --> 00:03:17.439
<v Speaker 2>look like. Researchers at the University of Hong Kong have

48
00:03:17.560 --> 00:03:22.240
<v Speaker 2>built an AI tool called cardiomic Score. It is an

49
00:03:22.280 --> 00:03:26.319
<v Speaker 2>AI powered blood test from a single blood draw. It

50
00:03:26.400 --> 00:03:30.240
<v Speaker 2>reads two thousand, nine hundred and twenty proteins and one

51
00:03:30.319 --> 00:03:33.840
<v Speaker 2>hundred and sixty eight metabolites in your blood stream, and

52
00:03:33.919 --> 00:03:36.879
<v Speaker 2>it uses a deep learning model to predict your risk

53
00:03:37.039 --> 00:03:42.520
<v Speaker 2>of six major cardiovascular diseases, including heart attack, stroke, and

54
00:03:42.639 --> 00:03:46.919
<v Speaker 2>heart failure up to fifteen years before any symptoms show up.

55
00:03:48.000 --> 00:03:52.960
<v Speaker 2>Let that sink in fifteen years of advance warning. The

56
00:03:53.039 --> 00:03:56.319
<v Speaker 2>model was trained and validated on a very large data

57
00:03:56.400 --> 00:04:00.199
<v Speaker 2>set from the UK Biobank, and it beat traditional risk

58
00:04:00.280 --> 00:04:04.639
<v Speaker 2>scores meaningfully in head to head comparisons. It is published

59
00:04:04.680 --> 00:04:08.719
<v Speaker 2>in a peer reviewed journal, not a press release. This

60
00:04:08.879 --> 00:04:13.360
<v Speaker 2>is real. Here is what happened, why it matters, and

61
00:04:13.479 --> 00:04:17.120
<v Speaker 2>what to do about it. What happened is that AI

62
00:04:17.439 --> 00:04:23.519
<v Speaker 2>just took cardiovascular medicine from reactive to genuinely predictive. Why

63
00:04:23.560 --> 00:04:26.399
<v Speaker 2>it matters for a small business owner is that The

64
00:04:26.399 --> 00:04:29.519
<v Speaker 2>cost of a preventable heart attack, whether it is you,

65
00:04:30.040 --> 00:04:33.920
<v Speaker 2>a co founder, a key employee, or a family member,

66
00:04:34.519 --> 00:04:38.439
<v Speaker 2>is measured in millions of dollars and sometimes in the

67
00:04:38.439 --> 00:04:43.319
<v Speaker 2>survival of the business. Once tests like cardiomic score reach

68
00:04:43.399 --> 00:04:47.600
<v Speaker 2>clinical rollout, they change the math on preventive care, on

69
00:04:47.720 --> 00:04:53.160
<v Speaker 2>group health premiums, and eventually on wellness program design. Now

70
00:04:53.240 --> 00:04:56.480
<v Speaker 2>the important part, which is the do this now list

71
00:04:57.600 --> 00:05:02.279
<v Speaker 2>Number one. If you offer group health benefits, put multiomic

72
00:05:02.360 --> 00:05:05.879
<v Speaker 2>biomarker screening on your radar for next year's plan review.

73
00:05:06.759 --> 00:05:10.519
<v Speaker 2>Ask your broker directly whether early adopter carriers will start

74
00:05:10.519 --> 00:05:15.600
<v Speaker 2>covering advanced risk stratification tests. The first carriers to do

75
00:05:15.720 --> 00:05:21.360
<v Speaker 2>so will price aggressively to win business. Number Two, for

76
00:05:21.480 --> 00:05:24.759
<v Speaker 2>yourself and any key person in your company whose absence

77
00:05:24.800 --> 00:05:29.000
<v Speaker 2>would hurt the business, Look into concierge or executive physical

78
00:05:29.040 --> 00:05:34.360
<v Speaker 2>programs that already offer proteomic and metabolomic panels. They are

79
00:05:34.399 --> 00:05:38.360
<v Speaker 2>not cheap, but they exist, and the delta versus a

80
00:05:38.399 --> 00:05:45.000
<v Speaker 2>standard physical is enormous. Number Three. Update your key person insurance.

81
00:05:45.920 --> 00:05:50.600
<v Speaker 2>If a fifteen year advance warning becomes clinically standard, underwriting

82
00:05:50.639 --> 00:05:54.639
<v Speaker 2>will shift and the businesses that document proactive screening for

83
00:05:54.680 --> 00:05:59.160
<v Speaker 2>their leadership will get better terms. The headline here is

84
00:05:59.199 --> 00:06:03.160
<v Speaker 2>a science store. The move for you is an insurance

85
00:06:03.199 --> 00:06:06.560
<v Speaker 2>and benefits move. Do not let this one sit in

86
00:06:06.600 --> 00:06:08.839
<v Speaker 2>the interesting but not urgent pile.

87
00:06:09.160 --> 00:06:13.279
<v Speaker 3>Now to Shanghai and a model launch that quietly changes

88
00:06:13.360 --> 00:06:18.560
<v Speaker 3>the competitive picture for anyone paying for frontier AI. At

89
00:06:18.600 --> 00:06:22.680
<v Speaker 3>the World Artificial Intelligence Conference in Shanghai this past week,

90
00:06:23.160 --> 00:06:28.040
<v Speaker 3>Ali Baba previewed its next flagship large language model. It

91
00:06:28.120 --> 00:06:32.720
<v Speaker 3>is called Quen three point eight Max Preview and it

92
00:06:32.800 --> 00:06:36.920
<v Speaker 3>is a two point four trillion parameter model for context

93
00:06:37.279 --> 00:06:40.519
<v Speaker 3>that is roughly ten times the size of the largest

94
00:06:40.600 --> 00:06:44.040
<v Speaker 3>model Ali Baba shipped just a year ago, and it

95
00:06:44.120 --> 00:06:47.360
<v Speaker 3>puts Quen in the same weight class as the biggest

96
00:06:47.480 --> 00:06:51.839
<v Speaker 3>closed models coming out of American labs. Ali Baba's Quen

97
00:06:51.959 --> 00:06:55.800
<v Speaker 3>team is publicly positioning the new model as trailing only

98
00:06:55.879 --> 00:07:00.600
<v Speaker 3>Anthropics top tier system, which they referred to as Claude

99
00:07:00.720 --> 00:07:05.720
<v Speaker 3>Fable five, and beating or matching essentially everyone else on

100
00:07:05.920 --> 00:07:12.160
<v Speaker 3>standard benchmarks. Now, benchmarks come with the usual caveats. Vendors

101
00:07:12.240 --> 00:07:15.800
<v Speaker 3>pick the tests that flatter them. Preview models are not

102
00:07:15.839 --> 00:07:20.839
<v Speaker 3>the same as generally available models, but the direction is unmistakable.

103
00:07:21.560 --> 00:07:25.279
<v Speaker 3>A high quality Chinese frontier model from a company that

104
00:07:25.319 --> 00:07:29.279
<v Speaker 3>has been aggressively pricing its AI services to gain market share,

105
00:07:29.800 --> 00:07:34.000
<v Speaker 3>is now genuinely in the top tier. So here is

106
00:07:34.040 --> 00:07:37.439
<v Speaker 3>what happened, why it matters, and what to do about it.

107
00:07:38.360 --> 00:07:41.639
<v Speaker 3>What happened is that the frontier AI market went from

108
00:07:41.680 --> 00:07:45.399
<v Speaker 3>a two or three horse race all American to a

109
00:07:45.480 --> 00:07:50.639
<v Speaker 3>four or five horse race with serious Chinese competition. Why

110
00:07:50.720 --> 00:07:55.360
<v Speaker 3>it matters is pricing. Every time a credible new frontier

111
00:07:55.439 --> 00:08:00.959
<v Speaker 3>model launches, per token prices on comparable models fall. Ali

112
00:08:01.000 --> 00:08:05.680
<v Speaker 3>Baba has historically undercut US competitors by fifty to eighty

113
00:08:05.720 --> 00:08:09.600
<v Speaker 3>percent on API pricing, and there is no reason to

114
00:08:09.639 --> 00:08:13.639
<v Speaker 3>think this generation will be different. Now the important part

115
00:08:14.439 --> 00:08:18.879
<v Speaker 3>Number one, ask your current AI vendor in writing for

116
00:08:18.959 --> 00:08:23.480
<v Speaker 3>their price roadmap through the next twelve months. Then quietly

117
00:08:24.000 --> 00:08:27.639
<v Speaker 3>run a technical evaluation of quen on the exact workloads

118
00:08:27.680 --> 00:08:34.399
<v Speaker 3>you use AI for today, coding, summarization, customer support, data extraction,

119
00:08:35.000 --> 00:08:38.039
<v Speaker 3>whatever it is. Do not do this because you want

120
00:08:38.039 --> 00:08:41.679
<v Speaker 3>to switch. Do it because leverage in your next renewal

121
00:08:41.720 --> 00:08:47.519
<v Speaker 3>conversation requires a real alternative sitting on your desk. Number two.

122
00:08:47.600 --> 00:08:50.840
<v Speaker 3>If your workload does not require top of the line reasoning,

123
00:08:51.279 --> 00:08:56.320
<v Speaker 3>and honestly, most SMB workloads do not. Price sensitive, open

124
00:08:56.320 --> 00:09:00.879
<v Speaker 3>weight or lower cost models are already good enough. The

125
00:09:00.919 --> 00:09:05.080
<v Speaker 3>gap between frontier and good enough is shrinking every quarter.

126
00:09:05.799 --> 00:09:09.159
<v Speaker 3>Every dollar you spend on frontier capability you did not

127
00:09:09.360 --> 00:09:12.600
<v Speaker 3>need is a dollar you did not spend on hiring,

128
00:09:12.919 --> 00:09:18.519
<v Speaker 3>marketing or product. Number three, be careful about data residency

129
00:09:18.759 --> 00:09:22.720
<v Speaker 3>and compliance if you operate in the US, the EU,

130
00:09:23.240 --> 00:09:27.240
<v Speaker 3>or in regulated industries. Using a Chinese model for anything

131
00:09:27.360 --> 00:09:31.159
<v Speaker 3>involving customer data is not a simple decision. Get your

132
00:09:31.200 --> 00:09:35.200
<v Speaker 3>privacy council involved before you pipe real data through any

133
00:09:35.399 --> 00:09:40.480
<v Speaker 3>non domestic model. Use these tools for internal experimentation and

134
00:09:40.720 --> 00:09:45.440
<v Speaker 3>non sensitive tasks. First. The takeaway is not that you

135
00:09:45.480 --> 00:09:50.360
<v Speaker 3>should abandon Anthropic or open AI. The takeaway is that

136
00:09:50.440 --> 00:09:53.679
<v Speaker 3>you now have real leverage in the negotiation, and you

137
00:09:53.720 --> 00:09:58.440
<v Speaker 3>should use it. This week's other big under the radar story,

138
00:09:59.320 --> 00:10:03.960
<v Speaker 3>Anthropic pushed enterprise customers from flat rate to full usage

139
00:10:04.000 --> 00:10:10.200
<v Speaker 3>based billing back in April, and GitHub Copilot followed weeks later. Meanwhile,

140
00:10:10.480 --> 00:10:14.919
<v Speaker 3>a KPMG survey of over two thousand executives found nearly

141
00:10:15.039 --> 00:10:19.000
<v Speaker 3>one in three has no idea where their AI costs

142
00:10:19.039 --> 00:10:24.000
<v Speaker 3>are actually coming from, and Uber publicly admitted it burned

143
00:10:24.080 --> 00:10:27.879
<v Speaker 3>through its entire twenty twenty six AI coding budget in

144
00:10:28.000 --> 00:10:32.639
<v Speaker 3>four months. The pattern is clear. The AI bill is

145
00:10:32.679 --> 00:10:36.240
<v Speaker 3>coming due, and the businesses that will thrive are the

146
00:10:36.240 --> 00:10:39.559
<v Speaker 3>ones that treat AI spending like a P and L

147
00:10:39.720 --> 00:10:44.200
<v Speaker 3>line item, not a magic wand which brings us to

148
00:10:44.240 --> 00:10:46.000
<v Speaker 3>the point of this whole show.

149
00:10:46.399 --> 00:10:50.320
<v Speaker 2>Here is something worth saying out loud. Every AI newsletter,

150
00:10:50.600 --> 00:10:54.679
<v Speaker 2>every AI podcast, every AI briefing you subscribe to will

151
00:10:54.679 --> 00:10:59.399
<v Speaker 2>tell you what happened This week. Current Ai launched a

152
00:10:59.440 --> 00:11:03.679
<v Speaker 2>blood test predicts heart attacks fifteen years out, Ali Baba

153
00:11:03.679 --> 00:11:08.399
<v Speaker 2>shipped to two point four trillion parameter model. Great. Now,

154
00:11:08.440 --> 00:11:13.639
<v Speaker 2>what Knowing what is happening is only part of the news. Frankly,

155
00:11:13.960 --> 00:11:17.159
<v Speaker 2>it is the easy part. The hard part, and the

156
00:11:17.200 --> 00:11:20.480
<v Speaker 2>actually valuable part is figuring out what to do about

157
00:11:20.480 --> 00:11:24.080
<v Speaker 2>it before your competitor does. That is what this show

158
00:11:24.120 --> 00:11:27.399
<v Speaker 2>tries to do every week. So let's put a bow

159
00:11:27.480 --> 00:11:29.679
<v Speaker 2>on it with the do this now list for the

160
00:11:29.720 --> 00:11:35.080
<v Speaker 2>week of July twentieth from the Current Ai story this week,

161
00:11:35.399 --> 00:11:39.320
<v Speaker 2>put one open source model on your evaluation list not

162
00:11:39.440 --> 00:11:43.039
<v Speaker 2>to switch to have a plan B and if you

163
00:11:43.080 --> 00:11:46.000
<v Speaker 2>operate in a non English speaking market, look at what

164
00:11:46.080 --> 00:11:49.399
<v Speaker 2>current AI is funding in your region. You may find

165
00:11:49.440 --> 00:11:52.120
<v Speaker 2>a model that fits your customers better than anything from

166
00:11:52.159 --> 00:11:58.240
<v Speaker 2>San Francisco from the Cardiomomic score story this month. Ask

167
00:11:58.279 --> 00:12:03.360
<v Speaker 2>your health benefits broker to questions when will advanced multi

168
00:12:03.399 --> 00:12:07.080
<v Speaker 2>omic screening be a covered benefit and is there an

169
00:12:07.120 --> 00:12:11.399
<v Speaker 2>early adopter carrier we should be talking to. Then, for

170
00:12:11.519 --> 00:12:14.960
<v Speaker 2>key people in your business, look into a proteomic panel

171
00:12:15.000 --> 00:12:19.320
<v Speaker 2>through a concierge health provider. Yes, it costs real money,

172
00:12:20.240 --> 00:12:24.039
<v Speaker 2>so does replacing your co founder From the Ali Baba

173
00:12:24.120 --> 00:12:28.639
<v Speaker 2>Quenn story. Before your next AI vendor renewal and no

174
00:12:28.759 --> 00:12:32.000
<v Speaker 2>later than ninety days from now, run a real technical

175
00:12:32.039 --> 00:12:35.879
<v Speaker 2>bake off of your workloads against at least one alternative model.

176
00:12:36.840 --> 00:12:41.279
<v Speaker 2>Document the results, bring them to the renewal conversation. That

177
00:12:41.480 --> 00:12:45.120
<v Speaker 2>single hour of preparation is worth more in dollar terms

178
00:12:45.279 --> 00:12:49.279
<v Speaker 2>than almost anything else on your calendar this quarter, and

179
00:12:49.320 --> 00:12:52.559
<v Speaker 2>the metapoint which applies to every AI story you will

180
00:12:52.600 --> 00:12:57.519
<v Speaker 2>read this year. When you hear a headline, ask three questions,

181
00:12:58.159 --> 00:13:01.879
<v Speaker 2>what actually happened, Why does it matter for a business

182
00:13:01.919 --> 00:13:06.679
<v Speaker 2>my size? And what specifically am I going to do

183
00:13:06.720 --> 00:13:11.320
<v Speaker 2>differently this week because of it. If you cannot answer

184
00:13:11.360 --> 00:13:14.600
<v Speaker 2>that third question, you have not really absorbed the news,

185
00:13:15.399 --> 00:13:19.080
<v Speaker 2>You have just consumed it. The businesses that will win

186
00:13:19.159 --> 00:13:21.679
<v Speaker 2>the next two years are not the ones that read

187
00:13:21.720 --> 00:13:25.120
<v Speaker 2>the most about AI. They're the ones that build the

188
00:13:25.159 --> 00:13:29.559
<v Speaker 2>tightest loop between hearing news and acting on it. Make

189
00:13:29.600 --> 00:13:33.799
<v Speaker 2>that loop tighter starting this week. That is the whole game.

190
00:13:35.000 --> 00:13:35.799
<v Speaker 2>See you next time.

191
00:13:37.080 --> 00:13:40.440
<v Speaker 1>That's it for this week's AI assembled the top AI

192
00:13:40.559 --> 00:13:44.120
<v Speaker 1>news for small business. If you found this valuable, share

193
00:13:44.159 --> 00:13:46.559
<v Speaker 1>it with a fellow business owner. And if you want

194
00:13:46.559 --> 00:13:49.600
<v Speaker 1>to know whether AI tools can find and recommend your business,

195
00:13:49.879 --> 00:13:55.600
<v Speaker 1>get your free AI website optimization report at freshmediaworks dot com.

196
00:13:55.639 --> 00:13:59.960
<v Speaker 1>Slash ai ready until next week, Keep innovating and keep

197
00:14:00.080 --> 00:14:00.480
<v Speaker 1>that real